At ULOGI, I managed analytics and reporting workloads across a 10+ TB PostgreSQL/TimescaleDB environment with truck, GPS, sensor, and IoT telemetry data. I designed compression, retention, and continuous aggregate policies that reduced daily location-data storage from approximately 10GB to 100MB while preserving historical data for analytics.
I also developed a multi-output ensemble regression system to predict ETA and fuel consumption from trip-level telemetry, and built dashboards for logistics operations. My work included custom GIS and routing solutions tailored to driver profiles, vehicle characteristics, operating country, and transported product.
At ULOGI, I built a customer-facing RAG backend with FastAPI, PostgreSQL, pgvector, LangGraph, and Ollama, with hybrid retrieval and voice capabilities. In my projects, I built a self-hosted RAG platform and an uncertainty-based hallucination classifier that achieved 0.988 F1 on HaluEval.
