Fractional CTO Case Studies

The Brief

Regeno builds land-management and compliance software for UK farmers, where the evidence that proves an agreement was met arrives as documents, photographs, audio and video rather than as tidy data. Two problems sat behind the AI work: knowledge the business already held that nothing could query, and a delivery backlog that no realistic amount of hiring was going to clear. Codenest was brought in to treat both as one platform problem rather than two initiatives.

The Build

The company brain makes Regeno's own records queryable and exposes them to agents over MCP, so that people and agents answer from what the business already knows rather than from whatever fits in a prompt. The AI factory is the delivery half. It takes a unit of work from where it already lives — a Linear ticket, a GitHub issue, a Slack message — and drives it to a merged pull request: typed workflow graphs, an isolated sandbox per run, bounded test and review loops, and a merge policy the operator sets. Every run leaves a ledger of its decisions, and writes back what it learned for the next one to recall.

What Is Being Built

• Queryable layer over the organisation's own records, exposed to agents over MCP
• Ticket in, reviewed and tested pull request out, running unattended
• Typed workflow graphs: bounded retry, review and merge gates rather than one long prompt
• A learning loop that ranks each run's lessons back into the next

Outcome figures go up once they are measured and the client has approved them. This engagement is still running, so there are none here yet.

Challenge

A UK payments provider needed to move off self-managed infrastructure and add new payment channels (Apple Pay, Google Pay), on a platform that had to keep processing transactions across Europe throughout.

Solution

Working inside the Opayo platform team from August 2019 to June 2026, Ankit led an AWS EKS migration with Kubernetes and Helm, GitOps workflows and Infrastructure as Code in Terraform, the move from a payments monolith to microservices, and Jenkins CI/CD pipelines running on Kubernetes.

Results

• Releases went from every two weeks to multiple times per day
• Apple Pay and Google Pay integrated as new payment channels
• Automated test gates in the pipeline, so a failing build stops before a release branch
• Monolith decomposed to microservices on AWS EKS, migrated in place

Challenge

AstraZeneca needed to replace manual spreadsheet-based drug delivery tracking with a modern web-based tool to improve efficiency and accelerate time-to-market for pharmaceutical products. The system required integration with existing workflows while maintaining regulatory compliance.

Solution

Ankit built a web application with REST APIs behind it, defining requirements with stakeholders and business analysts, deploying on Kubernetes with Docker and Jenkins as config-as-code, and putting the test suite in JUnit and Spock behind the pipeline.

Results

• Production-ready MVP delivered on Kubernetes infrastructure
• Spreadsheet-based tracking replaced by a single web tool
• Deployments automated end to end, with tests as a release gate
• Requirements agreed with stakeholders before each delivery increment

Challenge

A London-based HR-tech startup needed its social mentoring platform rebuilt to carry an enterprise client base. The system degraded badly beyond 5 concurrent users. Ankit joined as Lead Software Engineer, with the scope of a CTO, a year before Codenest existed.

Solution

Ankit led the rebuild: migrating from Neo4J to a hybrid MySQL/NoSQL primary data store (keeping Neo4J for AI/ML), microservices along domain boundaries, Infrastructure as Code on AWS, CI/CD pipelines, event-driven messaging over SQS, and containerising the stack.

Results

• Scaled from 5 to 1,000+ concurrent users
• Zero-downtime database migration
• Event-driven microservices replacing the original architecture
• CI/CD and Infrastructure as Code established from scratch