Rowell Camero

Designed and operated LLM-assisted automation pipelines to reduce repetitive financial admin work (data entry, document triage, follow-ups) with human-in-the-loop review. Built AI data workflows that extract structured fields via model-assisted mapping and validate accuracy/recall before production load. Shipped AI features into web UIs so staff can approve, correct, or override model outputs; encoded business rules into reliable transforms to support trustworthy production usage.

Built end-to-end document intelligence for extracting bill images into structured financial records for spend analysis using Python/Django and React. Implemented a two-stage extraction approach: Tesseract OCR followed by OpenAI LLM field mapping to normalize totals, taxes, and line items across messy layouts. Created a human review UI before final load and stored extracted attributes to support ongoing analytics and continuous AI quality improvement.

Architected TypeScript/Node.js microservices (Recorder, Feeder, Main, Analyzer, broker adapters) for continuous market and portfolio processing with AI-assisted trading insights. Built an Analyzer Engine combining rules with ML-assisted pattern recognition for near-real-time trading signals, risk metrics, and strategy features. Engineered real-time feature pipelines using WebSocket/Socket.IO, polyglot storage (PostgreSQL/Redis/MongoDB), and broker adapters for Interactive Brokers TWS and ETRADE. Exposed secured inference/data APIs (JWT/RBAC), delivered via Docker/Nginx, and implemented observability with ELK and Prometheus for latency, failures, and throughput.