Smart India Hackathon 2025 | Selected Project by Ministry of Earth Sciences

Built an AI-driven eDNA analysis pipeline for deep-sea biodiversity assessment, achieving 100× faster processing compared to traditional BLAST/DADA2 methods. Delivered a full-stack web application with real-time analysis dashboard, taxonomic classification trees, and biodiversity metrics visualization.

Team: Binary Beasts (Smart India Hackathon 2025)

🔗Website

Built an anomaly detection system for Food & Beverage batch production, analyzing 1500+ batches across 11 parameters. Implemented four specialized models with consensus voting, reducing false positives by 6-15% vs. single models.

Key Achievements: F1-score of 0.87 with 92% model agreement

🔗GitHub

Developed an adaptive, feedback-based AI tutor system using Meta's LLaMA-3.2-3B-Instruct with LangGraph multi-agent workflow for intelligent question generation, answer evaluation, and adaptive difficulty progression.

🔗GitHub |Hugging Face |Google Colab

Built a machine learning model to automatically generate detailed grading rubrics from given questions and sample answers. Fine-tuned on custom question-answer-rubric datasets for precise educational evaluation guidance.

🔗GitHub |Hugging Face

Built an AI-driven legal assistant specializing in Indian consumer law, traffic violations, and harassment cases. Implemented a multi-agent RAG system with query classification, intelligent semantic chunking, vector search (Qdrant), achieving 85%+ accuracy on legal queries.

Features: Real-time responses (<3s) • Privacy-focused • Session management • Production-ready deployment

🔗Live Bot |Website

Developed a Retrieval-Augmented Generation pipeline with Qdrant vector database for semantic search across multiple document formats (PDF, DOCX, CSV, JSON). Implemented hybrid search combining semantic embeddings and keyword matching.

🔗GitHub

Developed a pipeline to convert unstructured text into queryable Neo4j knowledge graphs, supporting multiple .txt files with intelligent chunking and type/relationship normalization.

🔗GitHub

Developed a secure AI-driven resume-job matching platform with multi-stage analysis using Gemini 2.5 and LangGraph workflows. Implemented AWS S3 presigned URLs for secure uploads with automatic 5-minute expiry.

🔗Website

WCHL Hackathon (Regional Round Qualifier, Asia)

Developed AI-powered service for India's blockchain-based digital governance platform, enabling document analysis & verification with OCR and text extraction, and grievance processing with sentiment analysis.

Achievement: Regional Round Qualification in Asia

🔗GitHub

First Hackathon Project: Developed an interactive web-based platform to train users in disaster preparedness and emergency response with AI chatbot powered by Gemini API.

🔗GitHub |Website

Global Game Jam 2025: Developed a first-person 3D puzzle-platformer during the 48-hour Global Game Jam 2025, showcasing rapid prototyping and game development skills.

Team: Devaansh Kushwaha, Hardik Bindal, Bhuvendra Singh

🔗Play Game

Developed a Python automation tool to extract text from images using Tesseract OCR, supporting batch processing with multi-language support (English & Hindi).

🔗GitHub

Conducted comprehensive Exploratory Data Analysis on popular datasets (Ames Housing, Iris Flower) covering data cleaning, feature analysis, and statistical visualization.

🔗GitHub