Engineered an end-to-end Sudoku Solver website where users can upload or input puzzles. Utilized OpenCV and AI-based OCR for digit extraction and a backtracking algorithm for solution generation. Achieved 92% digit recognition accuracy and solved 99% of puzzles within 2 seconds. Deployed via Docker on AWS EC2 with automated CI/CD using GitHub Actions.

Developed an AI-powered customer support chatbot using Streamlit and LangChain, integrating RAG for dynamic query resolution from internal documents and web sources, improving response accuracy by 35%. Implemented multi-LLM support (Groq Llama-3.1, OpenAI GPT-3.5, Google Gemini) with FAISS-based vector retrieval, reducing query latency by 40%.

Developed an interactive dashboard using Streamlit to visualize Olympic data trends, including medal tallies and athlete statistics. Implemented Python-based analytics for age distribution and performance insights across multiple sports. Deployed on Render for public access.

Developed a speech-based ML model for autism detection using ensemble learning (Random Forest and Naive Bayes) achieving 91% training and 87% validation accuracy. Processed MFCC audio features to improve classification by 25%. Collected and analyzed 120+ speech samples in collaboration with NGOs.