AI Engineer with 2.7+ years shipping LLM-backed features to production in TypeScript and Python. At Forest AI I build and own tool-calling AI agents run in production by AXA Group and Alteos, plus the evaluation harnesses and release gates that decide what ships — my evals raised agent accuracy from 78% to 95%. I also fine-tune small LLMs with LoRA and QLoRA.
Multi-tenant SaaS that turns any website or PDF into an embeddable RAG chatbot.
Shipped a multi-tenant SaaS turning any website or PDF into an embeddable RAG pipeline, with hybrid vector + keyword retrieval (reciprocal rank fusion).
Idempotent content-hash sync so only changed sources are re-embedded.
Protected ingestion against adversarial input through system design: SSRF-safe URL fetching and HMAC-signed visitor tokens.
More work
Fine-tuning experiments measured on held-out data, and contributions to open-source LLM tooling.
LLM fine-tuning
A 1.5B model fine-tuned to pull exact legal clauses out of contracts — without making quotes up.
Want to talk about a project or an opportunity? Email me at vdevwish12312@gmail.com.