[2025] WRITE Enso

- Python
- TypeScript
- React.js
- AWS
- Qdrant
- ClickHouse
- LLM Agents
- RAG
- Designed and built Enso's reporting/query engine from scratch, scaling into the platform's top endpoint (53K requests/week, 43x the next-busiest).
- Root-caused and fixed a critical latency bottleneck in the query engine, cutting p95 latency from 26s to under 1s and dashboard load from 34.7s to under 5s.
- Developed Enso's first AI agent from scratch: a tool-calling LLM agent generalized across 31 business object types via a 6-verb meta-toolkit, gated by staged write confirmations.
- Engineered a contract-intelligence pipeline (AWS Textract OCR + retrieval-augmented LLM parsing via Qdrant) with a self-improving few-shot loop.
- Led the proposal and solo development of a ClickHouse-based solution for high-volume event data, building the query-API layer across two distinct services.





