Emmanuel Okon

I’m a Freelance AI & Data Engineer focused on architecting production LLM systems that stay grounded, reliable, and measurable. I like building in phases—document ingestion, retrieval, serving, and evaluation—then tuning based on what fails in real deployments.

Recently, I architected and shipped a production RAG pipeline (NovaPay), using FAISS for retrieval, LangChain for orchestration, and FastAPI for serving, including confidence-score escalation. I also tuned prompts after watching live failures under active deployment and worked to block off-policy responses.

On the data & operations side, I engineered intelligence systems on 15M+ transactions, including an XGBoost agent approach and anomaly detection with Isolation Forest. I reduced CLV error from $111K to ~$49K MAE and delivered the workflows through FastAPI + Streamlit, while translating insights into operator-ready products.

I also built SME operations systems from zero (Supabase PostgreSQL schema, ingestion pipelines, order state tracking) to eliminate manual entry and deliver automated WhatsApp revenue visibility. My ML research background and end-to-end DS lifecycle—EDA, feature engineering, model training, evaluation—keep me practical, validated, and focused on business outcomes.