Created a mobile-first AI application for real-time crop disease detection aimed at assisting smallholder Indian farmers. Users upload crop images, and the app identifies diseases and recommends treatments.

Key features:

🌾 Supports classification of 38 crop diseases using a trained CNN

🧠 Integrated Generative AI to provide explanation and remedies in local languages

🌍 Designed with farmer-friendly UI for use in rural areas

📲 Runs on low-resource devices with optimized model performance

Tech stack: Python, TensorFlow, Streamlit, Hugging Face Transformers, TTS, OpenCV

eveloped a chatbot that assists users and advocates in analyzing legal cases by referencing previous judgments, evidence, and case law using RAG pipelines.

Key features:

⚖️ Answers legal queries based on uploaded documents

🔍 Uses vector DB + LLM for context retrieval

📚 References relevant past judgments and highlights supporting evidence

🧠 Supports contextual reasoning for legal advice and suggestions

Tech stack: LangChain, Mistral-7B, FAISS, Streamlit

Designed a deep learning-based system to detect counterfeit products from images. Trained on a labeled dataset of original and fake items to classify authenticity.

Key features:

🛍️ Image classification for real vs. fake products

📊 Model trained with CNNs and enhanced using data augmentation

🔍 Deployed prototype with UI for user image uploads

🌐 Future-ready for integration with e-commerce platforms

Tech stack: Python, TensorFlow/Keras, OpenCV, Flask/Streamlit

Built a real-time AI-powered interview simulation tool that analyzes user responses using voice, facial expressions, and LLM-based follow-up generation.

Key features:

🧠 Uses Whisper for real-time voice transcription

👁️ Detects facial emotions using computer vision

🤖 Integrates LLM agents (LLaMA 3.2) to generate personalized questions

📄 Generates interview performance reports with fluency feedback

💡 Built using Streamlit, OpenCV, LangChain, and RAG architecture

Tech stack: Python, Whisper, OpenCV, LLaMA 3.2, LangChain, Streamlit