Embedding model: all‑MiniLM‑L6‑v2
Vector store: NumPy cosine similarity (in‑memory)
Number of retrieved chunks per query: Top 5
I have been active on GitHub since 2022, presenting myself as a “BE IN COMPUTER SCIENCE AND ENGINEERING, AI/ML ENGINEER.” My profile showcases eight repositories spanning agricultural assistance, medical image classification, Retrieval‑Augmented Generation chatbots, bird classification, pure‑NumPy regression and credit‑risk pipelines, as well as churn and loan‑default prediction notebooks.
kisaan-sahayak: AI‑powered multilingual agricultural assistant for rural farmers in India. TypeScript, 2 stars. Last updated 2026-06.
I built Kisaan Sahayak as a multilingual AI‑powered agricultural assistant for Indian farmers. It runs on a modern front‑end stack and a serverless backend, and is publicly reachable at its live demo URL.
Live demo URL – https://kisaan-sahayak.web.app
- Multilingual support
Tech stack – React + TypeScript + Vite
Firebase (Firestore, Hosting, Auth)
Google Gemini AI API
Voice input support
- Crop disease diagnosis
- Livestock health guidance
- Agriculture news feed
- Community forum
- Field locator with maps
- Mandi price tracking
- Farm journal
skin-disease-classifier: AI model to classify 7 skin diseases using ResNet50 and fast.ai. Jupyter Notebook, 1 star. Last updated 2026-06.
rag-chatbot: RAG chatbot to chat with any PDF using Gemini 2.5 Flash, NumPy vector search, LangChain chunking, and Streamlit UI — with page‑level source citations. Python, 1 star, 1 fork. Last updated 2026-06.
fastai-bird-image-classifier: Bird vs forest image classifier built with fast.ai and transfer learning (ResNet). Jupyter Notebook, 1 star. Last updated 2026-06.
numpy-house-price-regression: End‑to‑end house‑price regressor in pure NumPy. Python. Last updated 2026-10.
credit‑risk‑scoring: LightGBM credit‑default risk pipeline with calibration, cost‑optimal threshold, SHAP explanations, fairness audit, FastAPI, Docker, CI. Python. Last updated 2026-10.
churn‑prediction: Jupyter Notebook. Last updated 2026-07.
loan‑default: Loan default prediction with logistic regression, decision tree, and random forest built from scratch in NumPy. Python. Last updated 2026-07.
Languages across the repos: Python (4), Jupyter Notebook (3), TypeScript (1).