I'm an AI/ML Engineer based in Hyderabad, India, specializing in building production-ready intelligent systems. My work spans retrieval-augmented generation (RAG), computer vision, medical imaging, and end-to-end ML pipelines—from data processing to deployment.

• 🔬 Deep Learning Research: Medical imaging, OCT retinal disease classification, liver segmentation
• 🤖 GenAI & LLMs: RAG applications, agentic systems, prompt engineering, fine-tuning (LoRA, full)
• 🏥 Healthcare AI: Diagnostic models, biomedical NLP, AI-driven healthcare solutions
• 📊 ML Engineering: Data pipelines, model evaluation, API development, cloud deployment

Data → Processing → ML/DL → Retrieval/LLM → Evaluation → API → Deployment

• IIT Madras Road Safety Hackathon Winner — ₹7.5 lakh grant for an AI-driven bike-safety enhancement system that identifies rash driving and potential falls, then alerts through a phone. Read the coverage

Languages & Core

AI / ML / Deep Learning

Data, Backend & APIs

DevOps, Cloud & Tools

• Architected a retrieval-augmented agent using LangChain with filtering, grounding, and multi-turn memory
• Integrated GPT-3.5 Turbo via API workflows — reduced average query cost by 10%
• Designed end-to-end ML pipelines: preprocessing, feature engineering, training, evaluation, and scikit-learn model deployment

• Built sEMG acquisition pipeline + LSTM for lower-limb abnormality detection (67% diagnostic accuracy)
• Pneumonia detection with MobileNet-V3 transfer learning — 94% accuracy, with Grad-CAM explainability
• Dockerized Django application deployed on AWS EC2

• Developed an MLP Mixture Model for retinal disease classification using Optical Coherence Tomography (OCT) images
• Achieved 98.4% accuracy, driving diagnostic precision in ophthalmology
• Tackled class imbalance through weighted loss functions, improving generalization and robustness for real-world healthcare deployment
• Applied advanced deep learning and computer vision methods to enhance diagnostic models

• Designed and optimized U-Net and Half U-Net architectures for precise liver segmentation
• Achieved 97.6% IoU and 9.4% Dice Coefficient, setting new standards in medical imaging accuracy
• Developed and tested custom loss functions to improve model performance
• Gained hands-on experience with AWS, Docker, and GPU-based cloud services (including NVIDIA A100 GPUs) to accelerate model training and deployment

RAG LangChain LangGraph ChromaDB Agentic AI LLM

A retrieval-powered FAQ assistant built for grounded Jupyter-related answers. The pipeline processes documentation, extracts FAQs, creates persistent embeddings, retrieves relevant context, and produces source-aware answers with agentic reasoning.

• Built an agentic RAG system with filtering, grounding, and multi-turn memory

• Used LangChain and LangGraph for orchestrating retrieval and generation workflows

• Implemented category-aware retrieval and citation-backed answers

  Live Demo

Django LLM Conversational AI Portfolio

An interactive portfolio website with an integrated conversational AI agent that provides information about projects, skills, and experience through natural language queries.

Live Demo

I write about AI/ML, deep learning, and healthcare AI on Medium.

🔗 Read my articles

Building AI systems that are useful, grounded, and measurable.

_{Made with ❤️ by Sai Krishna Chowdary Chundru}