Selected projects across legacy modernization, architecture, team recovery, and technical
strategy.
ArchitectureFeatured3 months
Two Critical Blockers, $2M Oversubscribed Seed Round
An early-stage enterprise AI application startup
Market: AI / SaaS
Problem
A viral post sent an early prototype under intense, unexpected load. With a funding close on the horizon, the team was blocked on two fronts: a codebase that had grown into multi-thousand-line Python scripts where safe changes were nearly impossible, and ad-hoc infrastructure running in tmux on a single EC2 instance that couldn't support what the product needed to become.
Approach
An AI-native engagement for an AI-native product. Worked in deep collaboration with frontier coding models throughout — compressing what would have been months of ramp-up and execution into a 3-month window where both blockers moved simultaneously. Restoring confidence in the codebase unlocked clearer thinking about product direction too, opening roadmap discussions that identified the two core ideas the team has run with through 2026.
•API service rebuilt from sprawling 10K-line scripts into a fully tested, verifiable FastAPI codebase — previously too large for IDEs to navigate, too risky to merge; now something the team could confidently move fast in
•Single over-resourced EC2 running ad-hoc scripts in tmux replaced with a containerized Terraform IaC blueprint — cost efficient auto-scaling, repeatable deployments, and tight GitHub integration across every layer
•CodeRabbit enrichment integrated alongside CI tooling, delivering targeted PR feedback that meaningfully tightened the code-to-production cycle
•Roadmap discussions unblocked two core product ideas the team has run with through 2026
Team RecoveryFeatured4 weeks initial, extended engagement
20-Person Team Turnaround in 4 Weeks
A healthcare technology company
Market: Healthcare
Problem
Brought in to augment development, but the real situation was more serious. Product was shipping non-meaningful changes to keep up appearances while the engineering team underneath had become disconnected, disheartened and disillusioned. There was no technical leadership — nobody holding the structure together or giving the team anything to orient around. A critical client relationship was already at risk from missed deliverables, and without intervention, there was no reason to expect that to change.
Approach
Sought out 1:1s proactively — what emerged was consistent: people were disheartened and felt unheard. Built alignment with the head of product over several conversations, earning a mandate to drive the team forward rather than simply being handed one. Established an engineering cycle process that brought the team around a shared structure. Worked with product to shape asks that matched what the client actually needed and what the team could realistically deliver. Made a point of listening to each team member individually — demonstrating their concerns were heard, and where possible, addressing them directly.
Outcomes
•20-person team reset and shipping productively within 4 weeks
•Critical client relationship repaired — high-value items delivered on time as promised
•Engineering cycle process established, giving the team structure and product realistic expectations for the first time
•Team morale recovered — from disheartened and disconnected to heard, aligned, and moving
Tech stack
Ruby on RailsPostgreSQL
Architecture16 months
250% Performance Improvement Before Black Friday
A two-sided marketplace connecting product creators with boutique sellers
Market: E-commerce / Marketplace
Problem
A Rails/React marketplace needed to scale for seasonal traffic spikes (Black Friday/Cyber Monday) while simultaneously building out engineering practices, observability, and data strategy from scratch. The platform team had no established standards, no SRE practice, and no data strategy — everything needed to move at once against a hard deadline.
Approach
Led platform architecture and the core platform team across the full breadth of what the role demanded. Established development processes and coding standards in the first three weeks, and introduced team leveling through lunch-and-learns and peer training sessions to raise the floor across the wider engineering org. Initialized and drove the buildout of SRE practices — monitoring, alerting, observability — training the team alongside. Ran multiple rounds of performance optimization against the Black Friday deadline. Deployed FiveTran and Snowflake to move critical queries off the live database. Revamped the hiring process to include more substantive technical challenges and a genuine 360-degree view of each candidate.
Outcomes
•250% system performance improvement — infrastructure costs reduced and seasonal spikes handled sustainably
•SRE practice established and team trained within 3 months
•Development standards and peer training programme in place within first 3 weeks
•Data strategy operational — critical queries migrated off production into FiveTran + Snowflake
•Hiring flow improved 30% through revamped process
Tech stack
Ruby on RailsReactPostgreSQLSnowflakeFiveTranDatadogAWS
DevEx6 months
From 20% Pipeline Completion to 30-Second Feedback
A major professional services firm's internal engineering team
Market: Professional Services
Problem
A critical internal project had a number of dysfunctional teams and a broken CI pipeline: an at-best 20% completion rate with 15-30 minute run times. Developers were operating without breathing room: PR review dwell time was killing velocity, and nobody had addressed the structural problems underneath.
Approach
Rapidly assessed the team's deficits and created a cross-functional plan using a document → discovery → act flow — giving the team structure and much-needed breathing space before pushing on delivery. Replaced the broken pipeline with efficient GitHub Actions, turning unreliable 30-minute waits into consistent 30-second feedback. Then led broader architectural decisions, resetting the team into established microservice patterns and decomposing a single MonoDatabase into per-service micro databases.
Outcomes
•CI feedback in 30 seconds (down from 15-30 minutes)
•Pipeline completion rate from 20% to consistent passes
•MonoDatabase decomposed into per-service micro databases
•Sustainable delivery cadence established
Tech stack
GitHub ActionsMicroservicesNode.js
Architecture4 weeks assessment, ongoing advisory
Existential Risk Assessment for a $100M ARR Business
A large direct-to-consumer e-commerce company ($100M ARR)
Market: E-commerce
Problem
The company identified an existential risk to their business and had started to address it, but needed an outside perspective to properly scope the problem and plan the response.
Approach
Performed an outside-in analysis of the situation. Compiled a detailed report and board deck identifying the exact shape of the problem. Then worked with the VPE to build action plans for resourcing, budget, and development cycles.
Outcomes
•Clear problem definition presented to leadership and board
•Replacement architecture plan with resourcing and budget
•Development cycle roadmap to eradicate the risk
Tech stack
Ruby on RailsAWS
Architecture8 weeks
40% Data Storage Reduction in 8 Weeks
A leading e-commerce design-to-consumer platform (Rails monolith)
Market: E-commerce
Problem
A lopsided database where most tables had mid-8-figure row counts, but two tables were in the billions. This created challenges in appropriately sizing and resourcing the data store, driving up costs and degrading performance.
Approach
Developed a series of queries to convert vertical semi-relational data into a de-normalized form on the parent record. Executed the migration while the platform continued to serve production traffic.
Outcomes
•40% storage reduction (50GB reclaimed)
•Significantly smaller database memory footprint
•Improved query performance on affected tables
Tech stack
Ruby on RailsPostgreSQL
Legacy Modernization6+ months
Rails 5 → 6 Migration Without Stopping the Train
A unicorn-sized startup with dozens of engineering teams
Market: Technology
Problem
A large Rails 5 application with significant home-grown functionality that needed careful migration to Rails 6 APIs — while dozens of teams continued to build and ship features without interruption.
Approach
Systematically identified and disconnected home-grown functionality, migrating it to Rails 6 native APIs. Designed the migration path to avoid blocking any team's feature work.
Outcomes
•50% of custom functionality migrated in first 6 weeks
•Zero interruption to feature teams throughout migration
Tech stack
Ruby on Rails
Tech Debt3 weeks assessment + implementation
20% Infrastructure Cost Reduction
An established consumer-focused publishing company
Market: Publishing / Media
Problem
A Node/React application that couldn't scale quickly and had high infrastructure costs, limiting the company's ability to grow.
Approach
Spent three weeks investigating the product and infrastructure, identifying areas of focus. Worked with in-house engineering teams to create and execute an action plan.
Outcomes
•20% reduction in ongoing infrastructure costs
•Pathway to faster build and deployment times
•Improved scaling capability
Tech stack
Node.jsReactAWS
Data Strategy3 weeks initial, 18 months total engagement
Event Pipeline and Audit Dashboard in 3 Weeks
A leading retail communications SaaS platform
Market: Retail Technology
Problem
Millions of events stored in the production database, impacting performance and making it difficult to gain operational insight into platform usage across major retail and grocery brands.
Approach
Led development and deployment of a Kinesis → Parquet → DynamoDB event pipeline, migrating millions of events off production. Established data management practices across the engineering team.
Outcomes
•Complete event pipeline operational in 3 weeks
•Audit dashboard delivered for customer visibility
•25% reduction in new customer onboarding time
•Production database load significantly reduced
Tech stack
Ruby on RailsAWS KinesisApache ParquetDynamoDBPostgreSQL
Legacy Modernization13 months
250K Lines of PHP → 16K Lines of Rails
A media technology company (TV network metadata services)
Market: Media / Entertainment
Problem
A bloated 250,000 KLOC PHP application serving as the media metadata API, running against a legacy database schema. The codebase was unmaintainable and couldn't support evolving business needs.
Approach
Led a tactical team to rebuild the API as a fully tested Rails application on the same legacy database and schema. Simultaneously delivered a robust SAML authentication scheme supporting federated auth across dozens of TV networks.
Outcomes
•250,000 KLOC reduced to 16,000 KLOC — fully tested
•Federated SAML authentication across dozens of networks
•Same legacy database, new maintainable codebase
Tech stack
Ruby on RailsPHPSAMLPostgreSQL
Architecture12 months
Rapid Prototyping for Agency Clients
A digital agency startup
Market: Agency / Professional Services
Problem
Agency needed future-looking prototype applications and proof-of-concepts to demonstrate strategic ideas for their clients, plus a production performance metric tool for a UK accounting firm.
Approach
Built multiple proof-of-concept applications. Authored SSO integration for the performance metric tool, enabling several thousand employees to use the system without prior account data.
Outcomes
•Multiple client-ready prototypes delivered
•SSO integration supporting thousands of employees
Tech stack
Node.jsReactSSO/SAML
Architecture3 months
From Prototype to Seed Funding
An early-stage sports trading card marketplace
Market: E-commerce / Startup
Problem
An unlaunched startup needed to rebuild their prototype product to secure funding and validate their market.
Approach
Selected Node.js/React/MongoDB stack for rapid iteration. Built a demonstrable product from scratch enabling initial sales and investor demos.