Rahul Singh

I’m a Machine Learning Engineer with 4+ years of production experience extending and shipping fuel analytics algorithms, building ML-backed detection systems, and developing multi-agent LLM applications. I focus on reliable, measurable improvements—from signal extraction to end-to-end deployment.

At Intangles, I extended and optimized a core DBSCAN + LOESS fuel event detection algorithm on OBD telemetry data, maintaining 95% alert accuracy across refill, theft, and data-loss events at scale. I built a sub-threshold fuel event detection system from scratch and a data-loss detection system that separates sensor dropout from genuine fuel events.

I also improved real-world performance by applying SMA-based smoothing to reduce false positives by 15% and refactoring a Node.js fuel analytics codebase to reclaim 30+ hours weekly by cutting support queries from 6 to 2 per day. I bring the same production mindset to LLM/RAG work—freshness-aware RAG with pgvector, multi-agent orchestration with LangGraph, and offline inference via llama.cpp—so insights are grounded and operationally dependable.