Hemant Kumar Chattar
About Candidate
I’m an engineer moving from embedded R&D into AI/LLM application engineering — building LLM-powered systems.
My foundation is real systems engineering. I was a key contributor to Chesto, India’s first tubeless digital stethoscope, working on BLE audio streaming and real-time signal processing for high-quality heart and lung sound transmission — building software that holds up under real constraints: latency, power, and hardware.
Now I’m applying that to AI. Recent projects, built from scratch to understand every layer:
• lore — a retrieval-augmented (RAG) assistant on the Claude API: local embeddings, hand-written cosine-similarity retrieval, and an evaluation harness.
• hawk — an image classifier converted (ONNX + INT8 quantization) and deployed to a Raspberry Pi, benchmarked on real hardware. Edge AI, measured.
• hound — an AI research agent: a multi-step tool-use loop (RAG over a vector DB + live web search) with an LLM-as-judge evaluation harness. Built from scratch, then re-implemented in LangGraph with multi-turn memory.
My edge: I build LLM applications and understand how to run models on constrained, real-world hardware — the on-device side.
Skills: Python · RAG · Agents · LangGraph · LLM APIs · Prompt Engineering · LLM Evaluation · ONNX / quantization · Embedded (BLE, firmware, C)
Location
Education
Work & Experience
Currently at LotiAI, I build generative-AI features for a video platform — multi-provider model aggregation, prompt-engineered scene consistency, and LLM→ElevenLabs narration.
•Built the BLE audio-streaming pipeline for Chesto — India's first tubeless digital stethoscope — delivering low-latency, high-fidelity transmission of heart and lung sounds for real-time auscultation.
•Developed real-time signal processing and embedded firmware (C/C++, RTOS) on resource-constrained hardware, tuning the audio path for diagnostic-grade sound quality and power efficiency.