Hey 👋, I'm Sanhith Reddy
I'm a final-year Data Science & AI student at IFHE Hyderabad, building intelligent systems at the intersection of Agentic AI, Retrieval-Augmented Generation (RAG ), and Computer Vision. I'm passionate about the mathematical foundations of ML and translating complex ideas into production-ready systems.
- 🎯 343 LeetCode problems solved (75.46% acceptance rate ) | 247+ GitHub contributions
- 🏆 Smart India Hackathon 2024: Team Lead for AI-driven Traffic Management (Top 5%)
- 🔍 Recently shipped: Videntia - 4-agent agentic system for video understanding (63% latency reduction, 94% accuracy)
- 🧠 Specialized in: Multi-agent orchestration (LangGraph), Hybrid RAG systems, Computer Vision pipelines
- 🔬 Deep-dived into fundamentals: Built Neural Networks from scratch using NumPy, implemented SVD-based image compression
- 💬 Let's talk about: Agentic AI, LLM orchestration, RAG systems, Computer Vision, and ML fundamentals
- 📙 📄 Resume
🎬 Videntia - Agentic Video Understanding System
- 4-agent LangGraph orchestration (Lead Detective, Retriever, Verifier, Report Generator )
- Hybrid retrieval: BM25 + Dense embeddings + Vision embeddings + Cross-encoder reranking
- 63% latency reduction (8 min → 3 min per hour) | 94% temporal logic accuracy
- Deployed on Hugging Face Spaces, Vercel, CLI
🤖 Agentic RAG Hybrid - Intelligent Research Assistant
- 3-strategy hybrid router: FAISS (private docs ) → Tavily (web search) → Chat fallback
- Query decomposition with pronoun resolution | Batch Q&A Lab (8 concurrent questions)
- 40% improvement in match accuracy over keyword-based methods
- Sub-2-second response times with GitHub Actions CI/CD
- Singular Value Decomposition achieving 75% file size reduction while preserving 90% SSIM
- Interactive rank selection with energy-based auto-tuning | Animated reconstruction visualization
- Optimized with thin SVD and float32 precision for real-time adjustments
- RAG-based semantic search achieving 40% accuracy improvement over keyword methods
- LLM-generated personalized explanations for job matches
- CNN-based detection with 92% test accuracy | Real-time video processing optimization
- Automated 10GB+ video preprocessing (cleaning, labeling, augmentation ) with 95% latency reduction
- Parallel processing pipeline using Python multiprocessing
- 🔄 Advanced multi-agent coordination patterns and state management
- 📈 Scaling RAG systems for production (handling 100GB+ knowledge bases )
- 🎬 Advanced computer vision techniques (3D reconstruction, pose estimation)
- ⚙️ MLOps and model deployment pipelines
I'm always excited to discuss Agentic AI, LLM orchestration, RAG systems, and Computer Vision. Feel free to reach out!
⚔️ Catching up on One Piece when I'm not training models....! ⚔️