Available for startup projects
I help early-stage startups ship LLM products, autonomous agents, and RAG pipelines — from prototype to production.
Available for new projects · US timezone · Replies within 24h
10+
tool integrations per agent
Live
production platform
Multi-LLM
OpenAI + Anthropic routing
Production agent platform where users deploy autonomous workers for research, coding, and task automation.
View live project →Real-time
analytics & backtesting
RAG
retrieval over financial data
PyTorch
forecasting models in prod
LLM-assisted strategy research and forecasting models powering a quant trading demo used by traders daily.
View live project →Schema-aware
structured JSON output
Eval-driven
prompt versioning & QA
API-first
predictable cost & latency
Schema-aware JSON extraction API replacing brittle regex pipelines with predictable latency and cost.
View live project →Streaming
real-time AI UI
RAG
retrieval over user data
Production
deployed on Vercel
End-to-end AI product with retrieval over user data and a streaming React experience.
View live project →“Roman shipped our agent orchestration and RAG pipeline fast. He thinks in products and production constraints, not just model demos.”
“Strong end-to-end delivery — from FastAPI and eval harnesses to the React UI. Exactly what we needed to get an LLM feature into production.”
“One of the few engineers who can own LLM architecture, backend APIs, and frontend UX in a single sprint. Reliable communicator.”
Ship AI-native features: chat, copilots, structured extraction, and workflow automation.
Design agent orchestration, tool routing, memory, and production guardrails.
Build retrieval pipelines, eval harnesses, and ML integrations with measurable quality.
End-to-end ownership from API and models to React/Next.js UI and deployment.