SethuramGautham Rajakumar

Built serverless AI voice agents on Azure (speech-to-text, NLP intent, n8n automations). Day-one failover and monitoring. Architected multi-agent systems: routing agent triages, specialists process, aggregation layer assembles outputs. Vendor-agnostic, upgradeable components. Researched RAG: benchmarked naive, advanced, and modular configs across chunking strategies and embedding models. Built Neo4j knowledge graphs for retrieval beyond vector search alone.

Deployed quantized TensorFlow Lite and ONNX models on Raspberry Pi and ESP32 for real-time OCR and anomaly detection, using INT8 quantization and hardware delegates to maintain inference speed on device with under 4GB RAM Built edge inference pipelines that processed camera and sensor data locally without cloud round-trips, achieving sub-100ms response times for time-critical industrial decisions

Curriculum covered machine learning, deep learning, computer vision, NLP, distributed systems, and data engineering. Gained hands-on experience building patented CNNs on Raspberry Pi, multi-agent systems with LangGraph, and RAG pipelines with Neo4j. Also certified in Anthropic MCP and Claude Code. Applied these skills in production at HGS Digital.