Anandhu p

I’m a final-year B.Tech AI & Data Science student with hands-on experience building production-grade AI backend systems. I specialise in agentic AI pipelines, Retrieval-Augmented Generation (RAG), LLM integration, real-time streaming systems, and REST API development with FastAPI—always aiming for modular, testable, and scalable architecture.

At Teqard Labs, I designed and deployed an AI-powered automatic hydroponic system by integrating ML-based sensor monitoring for real-time water and nutrient management using an IoT-to-cloud data flow. I also developed event-driven microcontroller firmware with finite-state machine control patterns, and embedded supervised ML models so inference directly influenced the hardware control loop.

On my CrisisLens project, I fine-tuned DistilBERT for 4-label multi-class disaster classification with severity scoring and sub-1.5s inference latency, then built a fully async event processing pipeline for end-to-end latency under 9 seconds. I integrated a FAISS-backed RAG layer before each Groq LLM call for explainability, and shipped a production-grade FastAPI REST backend with a live Streamlit ops dashboard, Docker containerisation, and automated verification.

My other work pushes reliability and explainability into the product: a MultiAgent AI Project Builder with a coordinator-worker FastAPI design, pluggable LLM adapter switching, cross-agent memory via ChromaDB/FAISS, and strong observability with pytest and CI. I’ve also built an edge wearable pipeline using ESP32 streaming + TensorFlow channel attention with SHAP and Groq-generated narratives, and a modular multimodal RAG ingestion system for PDFs using OCR/image captioning with a stage-swappable architecture.