Stack
What we build with.
The stack changes by problem. The rule does not: we choose tools that make the system maintainable after the initial excitement wears off.
AI / ML
- OpenAI API for mature product inference workflows
- Anthropic Claude where long-context reasoning is a better fit
- LangGraph and custom orchestration for agent state
- Whisper, Vapi, and ElevenLabs for voice-heavy systems
Infrastructure
- Vercel for frontend deployment and fast edge primitives
- AWS and Railway when infrastructure needs more control
- Supabase and Postgres when the data layer needs to stay legible
- Redis and queues where workflows need resilience under load
Development
- Next.js and TypeScript for product-grade web interfaces
- Python and FastAPI where backend AI services need speed and clarity
- Postgres and vector search when retrieval belongs inside the product
- Sentry and LangSmith for visibility when systems stop behaving