Applied AI

Applied AI is the direction I'm moving into. I build products on models, not research on them.

Current

RAG · Embeddings · LLM APIs · AI products

• Lexora: document intelligence and retrieval. Grounded Q&A with citations, per-user isolation, streaming, and an offline eval probe. FastAPI · PostgreSQL · Redis · FAISS
• EsenceLab: resume parsing and job matching behind a product UX. FastAPI · deterministic matching first, LLM optional

Where this goes

Inference · Evaluation · Retrieval quality

The pattern so far: retrieval quality is one input to an AI product. Auth, background processing, caching, evaluation, and tests are the system. Up next: measuring inference behavior, latency, and faithfulness on real workloads.