Case study
Work
Proof, not promises.
A system running in production today, demo builds that show how we approach common problems, and AI products we have built and shipped.
Free · 30 minutes · no obligation
Case study
In production today.
- documents it answers from
- 3,346
- searchable passages
- 26,894
- test passages from the right client, 0 from any other
- 20 / 20
- between checks for new files and email
- 3 min
Measured on the live system on 17 Sept 2026; the client-separation test was run on 29 July 2026.
Demo builds
How we'd approach common problems.
Sample builds that show how we put these systems together. They are demonstrations, not client work.
Lead to CRM automation
A sample workflow: when a website form comes in, Claude scores the lead, the contact is created or updated in HubSpot with no duplicates, the team gets a Slack alert and the lead gets a welcome email. If a step fails, it retries three times and then alerts the owner.
- n8n
- Claude
- HubSpot
- Slack
- Gmail
After-hours AI receptionist
A sample call flow: a voice agent answers the call, checks open slots in Google Calendar and books the appointment, logs the caller in GoHighLevel, texts a confirmation, and emails the owner a daily summary of every call.
- Vapi
- Google Calendar
- GoHighLevel
- SMS
Selected work
Products we've built and shipped.
AI sales agents, AI advisors that search course materials, and legal monitoring across all 50 states. Live products our team built and shipped.
Fastyr AI
Fastyr builds AI agents that work as members of a company’s team. We built the frontend of its AI sales platform from Figma designs: React and Next.js, GraphQL with Apollo, a Storybook component system, and complex API integrations.
The screen shows an earlier version of Fastyr’s site.
- React
- Next.js
- GraphQL
- AWS
OGE AI Wargaming
A wargaming platform for university courses. Students consult AI advisors that pair role-specific personas on Claude and OpenAI with search over their course materials, and negotiate in real-time chat. We built the full stack: the advisors, the retrieval pipeline, MCP tool integrations and the real-time layer.
- Claude
- OpenAI
- RAG
- MCP
- Django Channels
OGE AI Wargaming 3 screens
-
AI advisor settings -
RAG configuration -
Model and MCP server setup
OmniLegislation
Reads new bills, regulations and court opinions from all 50 states and federal sources around the clock, and sends plain-English digests of the ones that matter. We led the AI engineering: the monitoring engine, the summarization pipeline and the alerts.
The screens show an earlier version of the site.
- Python
- LLM pipelines
- Next.js
OmniLegislation 3 screens
-
Homepage, earlier version -
Topic search -
Activity by topic and jurisdiction
The screens show an earlier version of the site.
Your turn
What would you want off your plate?
Tell us about the work that keeps piling up. We'll tell you honestly whether a system would help.
Free · 30 minutes · no obligation