Products we’re proud to put our name on

Healthcare-costs.com

Data nobody could use became a goldmine of insight.

Before

Health insurers publish pricing data that is technically public, but practically unusable. Think hundreds of gigabytes of files per month, no index, no reference, no documentation.

After

We built a full-scale ETL pipeline on AWS that makes the whole corpus queryable. A healthcare research team got access to data they'd never previously been able to query to draw insight on pricing inconsistencies, insurer pricing variance, and health plan pricing variance.

Built

Traditional engineering teams estimated an 18-month buildout. We built our solution in 4 weeks: two weeks to build and two weeks to tune.

Visit healthcare-costs.com How we built it

Listbeaver.com

Inventory stopped being managed by hand.

Before

Sellers running ecommerce on more than one platform keep stock accurate the only way available to them: someone checks, someone updates, someone catches the mistake after the oversell.

After

Listbeaver syncs inventory across platforms so that job doesn't exist. One seller went from managing 100 listings on a single marketplace to over 1,000 across three.

Built

Syncs Instagram, Facebook, and TikTok marketplaces. Built on AWS with React and Supabase.

Visit listbeaver.com

Trenchcoat.io

Agent audit trails and cost sinks stopped being a black box.

Before

Teams are putting agents into production without any way to see how they make decisions, where they loop, or what's driving token cost. Improving agent reliability, tool usage, and token cost was a blind guess.

After

Trenchcoat taps into agent telemetry, like which agent spawned which subagent, what each skill activation cost, and where the spend really went. Agents can be optimized, redundancy can be eliminated, tool usage can be measured.

Results

• 80% lower token costs
• 50% smaller agent configuration footprint
• 25% fewer agents and agent tools to maintain

Built

Built on AWS with PostgreSQL, Next.js, Supabase, Claude API, and using OpenTelemetry.

Visit trenchcoat.io How we built it

BuildBot

Vendor lock-in became avoidable when investing in an Agent-building platform.

Before

Teams build up agents, skills, commands and hooks inside whichever AI coding platform they started on — and then can't move. The investment is real and it's stranded.

After

BuildBot makes each layer portable across platforms. Eleven export targets across the Claude, OpenAI, and Microsoft ecosystems. Over 100 Agent-building properties that have become interoperable.

Results

• 10x faster deploys — from hours of reconfiguring to instant export
• 200 agents and 50 skills in production across 10 teams
• 60% of users saw day-one benefit from eliminating migration cost

Built

Built with Bun, TypeScript, Next.js 16, React 19, Astro, Supabase, and Vercel.

Visit buildbot.me

AWS Resource Observability

Cloud spend stopped being a surprise at month end.

Before

Usage data exists. Nobody looks at it. Assembling the logs, building the report, and identifying a change is wishful thinking rather than a job that gets prioritised.

After

We built metering visualisation that makes consumption legible without anyone assembling anything. For a wellness non-profit it surfaced $400 a month going to configured resources nobody was using.

Built

Built with Next.js, FastAPI, AWS CDK, and ECS Fargate.

Healthcare Game of Life

Policy trade-offs you can watch instead of read about.

Before

Healthcare economics is taught in aggregates, which is exactly the level at which it stops being intuitive. The interesting behaviour is emergent, and aggregates hide emergence.

After

An agent-based simulation where the system-level outcome comes out of individual decisions you can change and re-run. Built for academic research, where researchers compare how design choices affect cost, fairness, and stability over time.

Built

Built with Python, FastAPI, Next.js, shadcn/ui, AWS CDK, ECS Fargate, and Playwright.