Selected work

Quietly powering ambitious teams.

Six case studies, six different industries. Each one shipped by a senior engineering team in 12 weeks or less.

AI · Operations
Northwind Logistics

Cut dispatch routing time by 38% with a custom AI model

Northwind ran 240+ delivery routes nightly through Excel macros. We built an end-to-end ML pipeline that ingests live traffic, driver availability, and SLA windows — and ships routes to dispatchers in under 8 seconds.

↓ 38% routing time99.4% on-time SLA8s end-to-end inference
PythonPyTorchFastAPIPostgres
Mobile · Health
PaceFit

4.9★ App Store ranking from a 12-week zero-to-launch build

PaceFit had a Figma file and a deadline. We shipped iOS + Android with offline-first workouts, HealthKit / Google Fit sync, and a coach handoff flow. Featured by Apple in week three.

12 weeks to App Store4.9★ avg rating70k MAUs in year one
React NativeSupabaseHealthKitStripe
Chatbot · SaaS
Foundry HR

12k support tickets/month deflected without lowering CSAT

Foundry was drowning in repetitive HR queries. We trained a 20-intent chatbot on 18 months of resolved tickets, with hard handoff rules to humans for anything sensitive. CSAT stayed flat while support headcount dropped 30%.

12k tickets/mo deflected↑ CSAT (kept above 4.6)↓ 30% support headcount
RasaSlack APIBigQueryLooker
Alexa · IoT
Hearth Energy

Top-10 Smart Home Alexa skill on launch day

Hearth wanted their smart thermostats voice-controllable. We shipped a certified Alexa Smart Home skill backed by Lambda, with multi-modal Echo Show support. Featured on the Alexa Skills Store in week one.

Top-10 Smart HomeCertification first-pass5★ avg in 4 weeks
Alexa Skills KitAWS LambdaDynamoDB
AI · Compliance
BrightLeaf Health

HIPAA-compliant RAG over 14k clinical protocols

A grounded AI assistant for nurses, with every answer cited to the source protocol. Zero hallucinations on our 240-case evaluation suite. Compliance signed off in one meeting.

0 hallucinations on eval14k docs indexedHIPAA + SOC 2 ready
ClaudepgvectorNext.jsAudit logging
ML · Forecasting
Mosaic Logistics

Demand forecast accuracy up 31% in one quarter

Mosaic had years of shipment data and a forecast that was useful only directionally. We rebuilt the model on cleaner features, deployed a daily retraining pipeline, and now the team plans warehouses six weeks out with confidence.

+31% accuracyDaily auto-retrainLive drift monitoring
ProphetAirflowMLflowGrafana

Have a project like one of these?

Book a 30-min call. We'll tell you whether we're a fit in the first ten minutes.

Book a free consultation