← All work
Founding CTO · Product Build
Inkra — turning conversations into finished work
Co-founding and building an AI platform that turns real conversations into completed, structured work for the non-technical teams running nonprofits — designed, built, and shipped to regulated standards.
Inkra · product hero image
In short
- Co-founded and built an AI platform that turns a single conversation — an intake call, a site visit, a case review — into notes, documents, and follow-up tasks, for people who aren't technical and shouldn't have to be.
- Owned the whole thing: product design, front end, back end, and the agent layer — not just the technical strategy.
- Built to regulated standards from the first commit: HIPAA-shaped workflows, SOC 2 posture, a differential-privacy layer, and a full audit trail.
The brief
Nonprofit program staff lose hours turning conversations into paperwork. The intake call has to become case notes; the site visit becomes a report; the review becomes a list of follow-ups. The people doing that work are program experts, not software users — and every extra tool is one more thing between them and the mission.
Inkra's job is to make the conversation the only input, and let the software produce everything downstream. The hard part isn't the demo — it's building something genuinely simple on the surface while handling sensitive data responsibly underneath, in a sector where a privacy mistake isn't an option.
Approach
I started from the person, not the model — designing the flow and interface around how program staff actually work, so the product hides the AI instead of showing it off. Then I built the pipeline behind it and engineered for regulated data from day one rather than retrofitting compliance later.
- Designed the conversational UI and end-to-end user flows for non-technical operators.
- Built the processing pipeline: transcription → LLM agents that extract, structure, and draft → human-reviewable output.
- Engineered data handling for compliance from the start — encryption, differential privacy, scoped auth, and an audit event schema.
- Kept the stack lean enough for a founding team to run, extend, and afford.
What I delivered
- A working product taken from zero to a usable platform — design through deployment.
- A conversational interface and user flows built for non-technical staff.
- A multi-agent processing pipeline with human-in-the-loop review gates.
- HIPAA-shaped data handling: Supabase Auth, a differential-privacy layer, and an audit event schema.
- Production infrastructure on AWS ECS Fargate — Next.js, FastAPI, PostgreSQL, Redis.
Outcomes
- ✓A platform non-technical staff can use without ever learning the AI underneath.
- ✓Compliance-grade data handling built in from day one, not bolted on before a deal.
- ✓A founding-team-friendly architecture that one technical leader can run and grow.
Stack & methods
Next.jsReactFastAPIPythonPostgreSQLRedisAWS ECS FargateSupabase AuthDifferential privacyLLM agentsHIPAA / SOC 2
Have a validated idea that needs a builder?
If you need someone who can design it, build it, and own the AI underneath — without a full-time hire — let's figure out if I'm the right fit.
Book a 30-min discovery call
See more work
