- Home
- Services: AppSec, DevOps & Observability
DevOps services that secure the entire pipeline
Our DevOps consulting services integrate security scanning, automated testing, and monitoring into a single pipeline, so every release moves from commit to production with its checks already passed.
DevOps challenges we help you solve
Security findings at pull request
Static analysis, dependency scanning, and secret detection run on every commit, surfacing issues to the developer who can fix them in minutes.
Deploys that take minutes, not meetings
CI/CD automation with environment promotion, approval gates, and rollback turns a release into a routine event any team member can trigger.
Signals ahead of support tickets
Metrics, logs, and traces reveal signs of degradation, with alerts focused on user impact.
Evidence generated by the pipeline
Capture your access controls, change records, and scan results automatically with each deployment, producing audit evidence as a byproduct of normal work.
How we approach DevOps, AppSec, and Observability
Security as a pipeline stage
Controls live inside the delivery workflow, where they run automatically, which keeps them from being skipped under deadline pressure.
Everything reproducible
Infrastructure, pipelines, and configuration are versioned as code, so any environment can be rebuilt and any change traced.
Observability tied to user experience
Service level objectives are defined around what users experience, such as latency and error rate, and alerts fire on those rather than on raw CPU.
Shared ownership of production
Developers see the dashboards and alerts for what they ship, which closes the gap between writing code and running it.
The practices behind the outcomes
Mantix, cloud infrastructure audit platform
Mantix checks cloud configurations against provider best practices, highlighting security gaps, resilience risks, and potential waste.
CI/CD Pipeline Engineering
Automate builds, tests, and deployments across GitHub Actions, GitLab, Azure Pipelines, and Jenkins.
AppSec Integration
SAST, DAST, dependency scanning, and container scanning are embedded as gates in the pipeline.
Infrastructure Automation
Provision and configure environments through Terraform and Helm, with guardrails defined as code.
Monitoring, Logging & Tracing
Configure Prometheus, Grafana, Datadog, and cloud-native tooling for end-to-end visibility.
Incident Response & SRE Practices
Establish on-call processes, runbooks, and post-incident reviews to shorten recovery and prevent repeats.
Related Services
Explore these additional services from our Platform & Cloud capability pillar.
Across the entire software development lifecycle, our platform engineering teams work within our AI-native Continuous Delivery Loop, eliminating handoffs and accelerating delivery.
We transform legacy systems into modern, cloud-native architectures to systematically reduce technical debt and unlock advanced functionality & usability.
We design and build native and cross-platform apps for iOS, Android, and embedded systems that consistently deliver experiences users trust on any device.
We break monolithic systems into independently scalable, production-hardened, integration-ready services that deploy faster and fail more safely.
We transition complex applications and systems to cloud platforms, reducing maintenance overhead while improving performance and reliability.
Built to pivot without losing velocity
Full product teams
Forward-deployed engineering
Co-sourced delivery
Rapid capacity deployment
Team augmentation
DevOps consultants who can streamline your entire delivery pipeline
World-class expertise covering every time zone and tech stack.
Practice Lead, DevOps & Automation
Senior technologist responsible for DevOps, Cloud, & Automation teams.
Technical Lead, DevOps & Automation
Senior technical leader specializing in DevOps, Cloud & Automation.
DevOps Architect
Senior technologist specializing in diverse DevSecOps platforms and practices.
Case Studies
BowFlex: Faster releases for connected fitness
More Case Studies
, hide_author_name_=false, module_id=224238135902, select_blog=224579243279, select_tag=healthtech-life-sciences}, child_css={}, css={}, id=module_179015393630755, label=Related Case Study Posts, module_id=224238135902, name=module_179015393630755, order=12, styles={}, type=module}, module_179120410606221={body={counter=[], logo_image={src=}, module_id=223627951216}, child_css={}, css={}, id=module_179120410606221, label=Case Study Posts Title, module_id=223627951216, name=module_179120410606221, order=2, styles={}, type=module}, module_179121160763349={body={module_id=223636303144, slider_image=[]}, child_css={}, css={}, id=module_179121160763349, label=fullwidth-image-slider, module_id=223636303144, name=module_179121160763349, order=8, styles={}, type=module}, module_179121390029660={body={discription=The challenge
An undocumented, fragmented codebase
Before building the API or writing new code, Coherent needed to identify the code currently driving the client’s companion app. With no product requirements documentation and four mismatched repositories to sort through, they had to cross-reference what was on old developer laptops against the live versions in the app stores.
No environment separation and rising cloud costs
The client’s systems and software were running in a single environment — without a clear separation between development, QA, and production — which increased the likelihood of issues like cascading errors. Additionally, cloud infrastructure spend was disproportionate to the size of the product, running close to $8,000 per month.
New hardware, new math
The client was bringing a new device to market that paired with the user’s smartphone and measured refractive error using patented optical technology. Coherent needed to build the backend logic that would take the client's proprietary formula, apply it to raw device measurements, and return an accurate, prescription-ready result.
, hide_section_=false, module_id=223640941568}, child_css={}, css={}, id=module_179121390029660, label=Case Study Content Section 1, module_id=223640941568, name=module_179121390029660, order=5, styles={}, type=module}, module_179121423312165={body={discription=Our approach
With little formal documentation to lean on, Coherent’s team relied on frequent, direct conversations with the client’s technical staff to confirm new features as they completed each step of the project:
- Codebase archaeology and consolidation: engineers worked through the four mismatched mobile repositories, cross-referenced them against the live app store build, and reconstructed the most current working codebase before making any changes.
- Environment separation: introduced dedicated development, QA, and production environments in place of the single, undifferentiated setup the client had been running.
- Cloud infrastructure cleanup: rebuilt cloud build pipelines and restructured infrastructure, improving reliability, while bringing monthly cloud costs down from roughly $8,000 to about $1,500, an 81% reduction.
- New device API: built a new backend API, on a Grails and Java/Spring stack, that takes raw measurements from the client's new refraction-testing hardware, applies the client's formula, and returns validated results ready for review by a licensed doctor.
The solution
Today, the platform supports a complete at-home vision-testing workflow. Users complete a guided assessment with the client's hardware, results are processed through Coherent-built backend logic, and outcomes are routed to a licensed eye care professional for a prescription. The underlying infrastructure now runs across separate dev, QA, and production environments at a fraction of its previous cloud cost, giving the client a stable foundation to build on as it brings new devices to market.
The impact
- 81% reduction in monthly cloud costs, from about $8,000 to about $1,500
- New device API delivered in about 9 months, supporting the client's newest hardware line
- New development, QA, and production environments to replace the client’s single setup, consolidating four mismatched mobile repositories into one working codebase
What's next
Through AKTIV, Coherent’s dedicated fitness and wellness practice, the company will continue to support the client’s ongoing platform enhancements across mobile, backend, DevOps, QA, and project management. Using the same consumer healthtech context that shaped the client’s rebuild, AKTIV's engineers will also ensure the client’s hardware, app, and backend continue to evolve together as the device line matures.
, hide_section_=false, module_id=223641064784}, child_css={}, css={}, id=module_179121475366271, label=case-study-content-section-4, module_id=223641064784, name=module_179121475366271, order=11, styles={}, type=module}, module_179121507763678={body={discription=[], module_id=223640945724}, child_css={}, css={}, id=module_179121507763678, label=case-study-three-column-card, module_id=223640945724, name=module_179121507763678, order=6, styles={}, type=module}, module_17921485850025={body={module_id=224578905090, quotes_description=, quotes_image={alt=, src=}, enable_to_hide_section=true}, child_css={}, css={}, id=module_17921485850025, label=case-study-full-width-quotes-block, module_id=224578905090, name=module_17921485850025, order=4, styles={}, type=module}, module_17923271166927={body={hide_show_module_=false, module_id=224756528181}, child_css={}, css={}, id=module_17923271166927, label=Case Study Video Banner Block, module_id=224756528181, name=module_17923271166927, order=10, styles={}, type=module}, blog_post_builder={body={sections=[{block_type=content_section, cs_group={content=The challenge
An undocumented, fragmented codebase
Before building the API or writing new code, Coherent needed to identify the code currently driving the client’s companion app. With no product requirements documentation and four mismatched repositories to sort through, they had to cross-reference what was on old developer laptops against the live versions in the app stores.
No environment separation and rising cloud costs
The client’s systems and software were running in a single environment — without a clear separation between development, QA, and production — which increased the likelihood of issues like cascading errors. Additionally, cloud infrastructure spend was disproportionate to the size of the product, running close to $8,000 per month.
New hardware, new math
The client was bringing a new device to market that paired with the user’s smartphone and measured refractive error using patented optical technology. Coherent needed to build the backend logic that would take the client's proprietary formula, apply it to raw device measurements, and return an accurate, prescription-ready result.
, top_section_space=none, enable_background_color=false, enable_top_border=false}}, {block_type=three_column, tc_group={cards=[81% reduction — In monthly cloud infrastructure costs, 9 months to deliver a new API — that supports the client’s at-home refraction-testing device, 3 separate environments created — for development, QA, and production to support validation and security], background_color={color=#f2f9f9, opacity=100}, light_teal_background_color={color=#83f0ff, opacity=100}, teal_dark_background_color={color=#006577, opacity=100}, midnight_grey_background_color={color=#253441, opacity=100}}}, {block_type=content_section, cs_group={content=Our approach
With little formal documentation to lean on, Coherent’s team relied on frequent, direct conversations with the client’s technical staff to confirm new features as they completed each step of the project:
- Codebase archaeology and consolidation: engineers worked through the four mismatched mobile repositories, cross-referenced them against the live app store build, and reconstructed the most current working codebase before making any changes.
- Environment separation: introduced dedicated development, QA, and production environments in place of the single, undifferentiated setup the client had been running.
- Cloud infrastructure cleanup: rebuilt cloud build pipelines and restructured infrastructure, improving reliability, while bringing monthly cloud costs down from roughly $8,000 to about $1,500, an 81% reduction.
- New device API: built a new backend API, on a Grails and Java/Spring stack, that takes raw measurements from the client's new refraction-testing hardware, applies the client's formula, and returns validated results ready for review by a licensed doctor.
The solution
Today, the platform supports a complete at-home vision-testing workflow. Users complete a guided assessment with the client's hardware, results are processed through Coherent-built backend logic, and outcomes are routed to a licensed eye care professional for a prescription. The underlying infrastructure now runs across separate dev, QA, and production environments at a fraction of its previous cloud cost, giving the client a stable foundation to build on as it brings new devices to market.
The impact
- 81% reduction in monthly cloud costs, from about $8,000 to about $1,500
- New device API delivered in about 9 months, supporting the client's newest hardware line
- New development, QA, and production environments to replace the client’s single setup, consolidating four mismatched mobile repositories into one working codebase
What's next
Through AKTIV, Coherent’s dedicated fitness and wellness practice, the company will continue to support the client’s ongoing platform enhancements across mobile, backend, DevOps, QA, and project management. Using the same consumer healthtech context that shaped the client’s rebuild, AKTIV's engineers will also ensure the client’s hardware, app, and backend continue to evolve together as the device line matures.
, top_section_space=none, enable_background_color=false, enable_top_border=false}}]}, child_css={}, css={}, id=blog_post_builder, label=Blog Post Builder, name=blog_post_builder, order=10.5, styles={}, type=module}}More Case Studies
, hide_author_name_=false, module_id=224238135902, select_blog=224579243279, select_tag=technology-saas}, child_css={}, css={}, id=module_179015393630755, label=Related Case Study Posts, module_id=224238135902, name=module_179015393630755, order=12, styles={}, type=module}, module_179120410606221={body={counter=[], logo_image={src=}, module_id=223627951216}, child_css={}, css={}, id=module_179120410606221, label=Case Study Posts Title, module_id=223627951216, name=module_179120410606221, order=2, styles={}, type=module}, module_179121160763349={body={module_id=223636303144, slider_image=[]}, child_css={}, css={}, id=module_179121160763349, label=fullwidth-image-slider, module_id=223636303144, name=module_179121160763349, order=8, styles={}, type=module}, module_179121390029660={body={discription=The challenge
Reviews often mention several characteristics of a product in the same text, each with its own opinion. An overall review score or sentiment label cannot capture that, and the client needed to know what customers thought about individual characteristics of each product.
The work covered hundreds of thousands of online product reviews, so the solution had to process customer feedback at scale.
, hide_section_=false, module_id=223640941568}, child_css={}, css={}, id=module_179121390029660, label=Case Study Content Section 1, module_id=223640941568, name=module_179121390029660, order=5, styles={}, type=module}, module_179121423312165={body={discription=Our approach
As the client's partner, Coherent Solutions engineered an NLP solution that analyzes online product reviews. Aspect-based sentiment analysis makes it possible to see which product characteristics customers react to, and how, across a large set of reviews.
Cloud engineering and scalable data processing allowed the solution to handle the large volume.
, hide_section_=false, module_id=223641064615}, child_css={}, css={}, id=module_179121423312165, label=Case Study Content Section 2, module_id=223641064615, name=module_179121423312165, order=7, styles={}, type=module}, module_179121449478468={body={discription=The solution
Coherent Solutions contributed:
- AI and machine learning engineering for a natural language processing solution built on aspect-based sentiment analysis
- Data processing designed for large volumes of customer feedback
- Cloud engineering supporting scalable data processing
The impact
The engagement produced the following results:
- Sentiment identified for specific product characteristics
- Product strengths and weaknesses identified from customer feedback
- Customer feedback processed at scale
What's next
How we worked here is how we work on every engagement. Whatever form your data takes, we start with the questions you need answered, select the right approach, and engineer the data processing and cloud deployment so the answers are ready to be used.
, hide_section_=false, module_id=223641064784}, child_css={}, css={}, id=module_179121475366271, label=case-study-content-section-4, module_id=223641064784, name=module_179121475366271, order=11, styles={}, type=module}, module_179121507763678={body={discription=[], module_id=223640945724}, child_css={}, css={}, id=module_179121507763678, label=case-study-three-column-card, module_id=223640945724, name=module_179121507763678, order=6, styles={}, type=module}, module_17921485850025={body={module_id=224578905090, quotes_description=, quotes_image={alt=, src=}, enable_to_hide_section=true}, child_css={}, css={}, id=module_17921485850025, label=case-study-full-width-quotes-block, module_id=224578905090, name=module_17921485850025, order=4, styles={}, type=module}, module_17923271166927={body={hide_show_module_=false, module_id=224756528181}, child_css={}, css={}, id=module_17923271166927, label=Case Study Video Banner Block, module_id=224756528181, name=module_17923271166927, order=10, styles={}, type=module}, blog_post_builder={body={sections=[{block_type=content_section, cs_group={content=The challenge
Reviews often mention several characteristics of a product in the same text, each with its own opinion. An overall review score or sentiment label cannot capture that, and the client needed to know what customers thought about individual characteristics of each product.
The work covered hundreds of thousands of online product reviews, so the solution had to process customer feedback at scale.
, top_section_space=none, enable_background_color=false, enable_top_border=false}}, {block_type=three_column, tc_group={cards=[Reviews read at scale — hundreds of thousands of online product reviews analyzed, Feature-level opinions, not just star ratings — sentiment tied to specific product characteristics, Cloud engineering at scale — scalable data processing supported by cloud engineering], background_color={color=#f2f9f9, opacity=100}, light_teal_background_color={color=#83f0ff, opacity=100}, teal_dark_background_color={color=#006577, opacity=100}, midnight_grey_background_color={color=#253441, opacity=100}}}, {block_type=content_section, cs_group={content=Our approach
As the client's partner, Coherent Solutions engineered an NLP solution that analyzes online product reviews. Aspect-based sentiment analysis makes it possible to see which product characteristics customers react to, and how, across a large set of reviews.
Cloud engineering and scalable data processing allowed the solution to handle the large volume.
, top_section_space=pt40, enable_background_color=false, enable_top_border=false}}, {block_type=content_section, cs_group={content=The solution
Coherent Solutions contributed:
- AI and machine learning engineering for a natural language processing solution built on aspect-based sentiment analysis
- Data processing designed for large volumes of customer feedback
- Cloud engineering supporting scalable data processing
The impact
The engagement produced the following results:
- Sentiment identified for specific product characteristics
- Product strengths and weaknesses identified from customer feedback
- Customer feedback processed at scale
What's next
How we worked here is how we work on every engagement. Whatever form your data takes, we start with the questions you need answered, select the right approach, and engineer the data processing and cloud deployment so the answers are ready to be used.
, top_section_space=none, enable_background_color=false, enable_top_border=false}}]}, child_css={}, css={}, id=blog_post_builder, label=Blog Post Builder, name=blog_post_builder, order=10.5, styles={}, type=module}}More Case Studies
, hide_author_name_=false, module_id=224238135902, select_blog=224579243279, select_tag=financial-services}, child_css={}, css={}, id=module_179015393630755, label=Related Case Study Posts, module_id=224238135902, name=module_179015393630755, order=12, styles={}, type=module}, module_179120410606221={body={counter=[], logo_image={src=}, module_id=223627951216}, child_css={}, css={}, id=module_179120410606221, label=Case Study Posts Title, module_id=223627951216, name=module_179120410606221, order=2, styles={}, type=module}, module_179121160763349={body={module_id=223636303144, slider_image=[]}, child_css={}, css={}, id=module_179121160763349, label=fullwidth-image-slider, module_id=223636303144, name=module_179121160763349, order=8, styles={}, type=module}, module_179121390029660={body={discription=The challenge
The lender had outgrown the way it built software. Growth had added systems, teams and integrations faster than it added the documentation and standards to coordinate them, and leadership recognized the company could no longer run the way it had at a quarter of the size.
Five problems shaped the engineering agenda.
-
Two CRMs, one business. A Salesforce implementation and a PHP platform built more than a decade earlier both ran in production. For older products, sales sat in Salesforce while underwriting, collections and payments stayed in the legacy platform.
-
Manual underwriting and slow decisioning. Prequalification and approval depended on manual steps and on data pulled individually from each credit, identity and document provider.
-
Releases that rolled back. Testing happened in UAT, defects surfaced in production, and rollbacks followed. On live funding and payment flows, each one carried operational cost.
-
No shared record of how anything worked. Onboarding, handover and incident response all depended on the people who happened to remember.
-
A bought platform that did not fit. Loan origination ran on an off-the-shelf digital lending product, customized for the lender by the vendor that built it. The architecture did not match how the lender works, the application carried a large set of features the business never used, and support and hosting fees continued regardles
Our approach
Coherent started inside the lender's existing teams and moved into ownership as the engagement proved out. The lender's engineering managers ran delivery at the outset and brought Coherent in for capacity. That changed, stream by stream, into accountability for outcomes.
Taking loan origination out of the vendor's hands
The vendor that built the loan origination platform had customized it for the lender and continued to develop and support it. The lender asked Coherent to move that development onto our team.
The vendor had no reason to move quickly. Its own processes governed how requirements were confirmed and how work shipped, and its contract paid it for the work being transferred away. For close to a year the two teams ran in parallel, at first on the same stories, with scope released to Coherent's side only as far as the vendor's position allowed. Our delivery lead worked inside that arrangement, learning how the implementation had been built and why, while the team put its engineers through the certifications the platform requires. When the vendor's contract ended, the whole of it sat with Coherent.
We own development and releasing today, with three developers and a QA engineer on the stream. DevOps sits on the lender's side with a Coherent engineer inside that team, and the lender keeps approval rights over what ships. The lender has been rebuilding how it plans and releases, so the team now works to a different process than the one the vendor ran.
Scoping a checkout the lender would own
The lender is now scoping a checkout process of its own, to replace the part of the vendor platform it actually uses. The monthly licensing and hosting bill covers the whole product, not the portion the business runs on.
The lender has started making this concrete. It brought in a product owner dedicated to the rebuild, and it is planning to grow the team so one group supports the existing process while another builds the new checkout.
Consolidating on Salesforce while keeping legacy supportable
Salesforce is the lender's strategic system of record, and five teams work on it. The newest product was built Salesforce-only from the start, which set the direction for what follows. A new partner portal, close to release, connects directly to Salesforce and bypasses the older partner API.
The legacy PHP systems still carry core processes, so the work there is about keeping them safe to run. Coherent upgraded the partner API to a supported PHP version, and the older platform continues to serve the products that depend on it while the lender decides how to retire it.
Simplifying integration and data
Coherent's integration work reduced the number of connections the lender has to own.
- An event bus carries traffic between internal systems and out to third parties on a publish and subscribe standard. A payment event reaches every subscribed system without bespoke wiring.
- Credit, identity, document intelligence and background data now arrive through a single gateway rather than a direct integration per provider, from sources including Experian, Plaid, Ocrolus and LexisNexis.
- Mambu orchestrates accounts and payments between Salesforce and the payment systems, including bank responses.
- Snowflake consolidates the databases behind every system, and a dedicated team builds curated views. A team needing data from several sources reads one view instead of many databases.
Establishing QA and release standards
Coherent's QA lead owned quality across the lender's entire delivery organization, internal and external teams alike, for roughly 18 months, until the lender hired a permanent QA director. The work concentrated on the rollback problem, covering how releases are tested, what has to be true before one ships, and which standards apply to every team rather than to one. Scrum practice was strengthened alongside it.
Putting AI into the delivery workflow
Developers work in Cursor, and QA adopted it for test case authoring. The lender approved the tooling and provided the licenses, which matters for a lender with compliance exposure on how code and test artifacts are produced.
, hide_section_=false, module_id=223641064615}, child_css={}, css={}, id=module_179121423312165, label=Case Study Content Section 2, module_id=223641064615, name=module_179121423312165, order=7, styles={}, type=module}, module_179121449478468={body={discription=The solution
The lender now has a delivery organization with standards and one strategic direction for its platform. The application path runs end to end. An application arrives from direct sales or a partner, data on the business and its owner returns through the single gateway, the backend calculates a risk score, and a go or no-go decision follows. Approved merchants receive offers, negotiate terms and sign contracts, funding runs through the lender's bank agreements, and collections tracks payments against schedule.
Three surfaces sit around that path.
- A merchant portal where borrowers see account history and payments, request a promise to pay, and generate their own monthly statements.
- A new partner portal wired directly to Salesforce, replacing an older portal and the API hop behind it.
- A case manager application that receives merchant-run checkout submissions and approves or declines them.
The organization around the software changed as much as the software did. QA and release standards are documented and shared, and loan origination has an accountable delivery owner inside Coherent rather than at a vendor.
The impact
This work left the lender with capabilities it did not have before, and a transaction that depended on them.
- Business readiness. The lender's technology leadership spent this period preparing the company for a transaction with a private-equity buyer. That transaction completed, and Coherent's engineering and process work supported the readiness behind it.
- Delivery no longer runs on a vendor's timetable. Loan origination changes ship on the lender's schedule, released by a certified team accountable for the outcome rather than for a support contract.
- Releases stopped being a risk event. Rollback-free releases are the explicit target across every delivery team, measured against one documented standard where there previously was none.
- Adding a data provider is a one-place change. A new credit, identity or document source no longer means another point-to-point integration to own and maintain.
- Answering a data question no longer requires knowing the estate. Curated views replaced the need to track which of many systems held which field.
What's next
The checkout rebuild is the next substantial piece of work, and it is not ready to start. The lender wants development underway early next year, which means requirements and a first set of stories have to exist before this year ends. The team has seen a two-page outline so far, which means the schedule depends on analysis that has barely started.
The migration off the legacy platform is planned rather than underway. The lender is weighing the cost of moving the remaining products into Salesforce against the cost of upgrading the legacy platform's PHP version, and core processes still depend on the old platform. Meanwhile the new partner portal is close to release and AI-assisted delivery is expanding from development into QA.
This engagement shows where modernization tends to lead at a company that grew faster than its documentation. Two years in, the question in front of the business is no longer how to modernize a platform it bought. It is which parts of that platform it should own outright.
, hide_section_=false, module_id=223641064784}, child_css={}, css={}, id=module_179121475366271, label=case-study-content-section-4, module_id=223641064784, name=module_179121475366271, order=11, styles={}, type=module}, module_179121507763678={body={discription=[], module_id=223640945724}, child_css={}, css={}, id=module_179121507763678, label=case-study-three-column-card, module_id=223640945724, name=module_179121507763678, order=6, styles={}, type=module}, module_17921485850025={body={module_id=224578905090, quotes_description=, quotes_image={alt=, src=}, enable_to_hide_section=true}, child_css={}, css={}, id=module_17921485850025, label=case-study-full-width-quotes-block, module_id=224578905090, name=module_17921485850025, order=4, styles={}, type=module}, module_17923271166927={body={hide_show_module_=false, module_id=224756528181}, child_css={}, css={}, id=module_17923271166927, label=Case Study Video Banner Block, module_id=224756528181, name=module_17923271166927, order=10, styles={}, type=module}, blog_post_builder={body={sections=[{block_type=content_section, cs_group={content=The challenge
The lender had outgrown the way it built software. Growth had added systems, teams and integrations faster than it added the documentation and standards to coordinate them, and leadership recognized the company could no longer run the way it had at a quarter of the size.
Five problems shaped the engineering agenda.
-
Two CRMs, one business. A Salesforce implementation and a PHP platform built more than a decade earlier both ran in production. For older products, sales sat in Salesforce while underwriting, collections and payments stayed in the legacy platform.
-
Manual underwriting and slow decisioning. Prequalification and approval depended on manual steps and on data pulled individually from each credit, identity and document provider.
-
Releases that rolled back. Testing happened in UAT, defects surfaced in production, and rollbacks followed. On live funding and payment flows, each one carried operational cost.
-
No shared record of how anything worked. Onboarding, handover and incident response all depended on the people who happened to remember.
-
A bought platform that did not fit. Loan origination ran on an off-the-shelf digital lending product, customized for the lender by the vendor that built it. The architecture did not match how the lender works, the application carried a large set of features the business never used, and support and hosting fees continued regardles
Our approach
Coherent started inside the lender's existing teams and moved into ownership as the engagement proved out. The lender's engineering managers ran delivery at the outset and brought Coherent in for capacity. That changed, stream by stream, into accountability for outcomes.
Taking loan origination out of the vendor's hands
The vendor that built the loan origination platform had customized it for the lender and continued to develop and support it. The lender asked Coherent to move that development onto our team.
The vendor had no reason to move quickly. Its own processes governed how requirements were confirmed and how work shipped, and its contract paid it for the work being transferred away. For close to a year the two teams ran in parallel, at first on the same stories, with scope released to Coherent's side only as far as the vendor's position allowed. Our delivery lead worked inside that arrangement, learning how the implementation had been built and why, while the team put its engineers through the certifications the platform requires. When the vendor's contract ended, the whole of it sat with Coherent.
We own development and releasing today, with three developers and a QA engineer on the stream. DevOps sits on the lender's side with a Coherent engineer inside that team, and the lender keeps approval rights over what ships. The lender has been rebuilding how it plans and releases, so the team now works to a different process than the one the vendor ran.
Scoping a checkout the lender would own
The lender is now scoping a checkout process of its own, to replace the part of the vendor platform it actually uses. The monthly licensing and hosting bill covers the whole product, not the portion the business runs on.
The lender has started making this concrete. It brought in a product owner dedicated to the rebuild, and it is planning to grow the team so one group supports the existing process while another builds the new checkout.
Consolidating on Salesforce while keeping legacy supportable
Salesforce is the lender's strategic system of record, and five teams work on it. The newest product was built Salesforce-only from the start, which set the direction for what follows. A new partner portal, close to release, connects directly to Salesforce and bypasses the older partner API.
The legacy PHP systems still carry core processes, so the work there is about keeping them safe to run. Coherent upgraded the partner API to a supported PHP version, and the older platform continues to serve the products that depend on it while the lender decides how to retire it.
Simplifying integration and data
Coherent's integration work reduced the number of connections the lender has to own.
- An event bus carries traffic between internal systems and out to third parties on a publish and subscribe standard. A payment event reaches every subscribed system without bespoke wiring.
- Credit, identity, document intelligence and background data now arrive through a single gateway rather than a direct integration per provider, from sources including Experian, Plaid, Ocrolus and LexisNexis.
- Mambu orchestrates accounts and payments between Salesforce and the payment systems, including bank responses.
- Snowflake consolidates the databases behind every system, and a dedicated team builds curated views. A team needing data from several sources reads one view instead of many databases.
Establishing QA and release standards
Coherent's QA lead owned quality across the lender's entire delivery organization, internal and external teams alike, for roughly 18 months, until the lender hired a permanent QA director. The work concentrated on the rollback problem, covering how releases are tested, what has to be true before one ships, and which standards apply to every team rather than to one. Scrum practice was strengthened alongside it.
Putting AI into the delivery workflow
Developers work in Cursor, and QA adopted it for test case authoring. The lender approved the tooling and provided the licenses, which matters for a lender with compliance exposure on how code and test artifacts are produced.
, top_section_space=pt40, enable_background_color=false, enable_top_border=false}}, {block_type=content_section, cs_group={content=The solution
The lender now has a delivery organization with standards and one strategic direction for its platform. The application path runs end to end. An application arrives from direct sales or a partner, data on the business and its owner returns through the single gateway, the backend calculates a risk score, and a go or no-go decision follows. Approved merchants receive offers, negotiate terms and sign contracts, funding runs through the lender's bank agreements, and collections tracks payments against schedule.
Three surfaces sit around that path.
- A merchant portal where borrowers see account history and payments, request a promise to pay, and generate their own monthly statements.
- A new partner portal wired directly to Salesforce, replacing an older portal and the API hop behind it.
- A case manager application that receives merchant-run checkout submissions and approves or declines them.
The organization around the software changed as much as the software did. QA and release standards are documented and shared, and loan origination has an accountable delivery owner inside Coherent rather than at a vendor.
The impact
This work left the lender with capabilities it did not have before, and a transaction that depended on them.
- Business readiness. The lender's technology leadership spent this period preparing the company for a transaction with a private-equity buyer. That transaction completed, and Coherent's engineering and process work supported the readiness behind it.
- Delivery no longer runs on a vendor's timetable. Loan origination changes ship on the lender's schedule, released by a certified team accountable for the outcome rather than for a support contract.
- Releases stopped being a risk event. Rollback-free releases are the explicit target across every delivery team, measured against one documented standard where there previously was none.
- Adding a data provider is a one-place change. A new credit, identity or document source no longer means another point-to-point integration to own and maintain.
- Answering a data question no longer requires knowing the estate. Curated views replaced the need to track which of many systems held which field.
What's next
The checkout rebuild is the next substantial piece of work, and it is not ready to start. The lender wants development underway early next year, which means requirements and a first set of stories have to exist before this year ends. The team has seen a two-page outline so far, which means the schedule depends on analysis that has barely started.
The migration off the legacy platform is planned rather than underway. The lender is weighing the cost of moving the remaining products into Salesforce against the cost of upgrading the legacy platform's PHP version, and core processes still depend on the old platform. Meanwhile the new partner portal is close to release and AI-assisted delivery is expanding from development into QA.
This engagement shows where modernization tends to lead at a company that grew faster than its documentation. Two years in, the question in front of the business is no longer how to modernize a platform it bought. It is which parts of that platform it should own outright.
, top_section_space=none, enable_background_color=false, enable_top_border=false}}]}, child_css={}, css={}, id=blog_post_builder, label=Blog Post Builder, name=blog_post_builder, order=10.5, styles={}, type=module}}Ready to optimize your infrastructure?
Relevant
Questions
CI integrates and tests code changes. CD keeps changes ready for release, with approval where needed; continuous deployment releases them automatically. Production automation depends on test coverage and risk tolerance.
Make the pipeline generate the evidence, meaning approval records, scan results, and access logs captured automatically per deployment. Audits become a reporting exercise rather than a document hunt.
Monitoring tells you a known metric crossed a threshold. Observability lets you ask new questions about system behavior from logs, metrics, and traces when something unexpected happens.
Security checks run at commit and build time, so developers see findings while the code is fresh. Deeper security testing complements these checks before release. The practical result is cheaper fixes and fewer release-blocking surprises.
Run fast checks like secret detection and linting on every commit, run heavier scans like DAST on merges or nightly, and tune rules to suppress noise. Review scan times and findings regularly, and tune configuration and scheduling to keep feedback fast while maintaining security coverage.
Track deployment frequency, change lead time, change failure rate, and time to restore. Improvements in those four metrics correlate with both delivery speed and stability.