Portfolio
Explore my portfolio of GenAI, machine learning, full-stack AI, and data analytics projects. Each project demonstrates end-to-end development across LLM pipelines, ML modeling, and production deployments.
AI-Powered Automated Testing Agent
LangGraph orchestrates multi-step Gemini reasoning to convert plain-English test cases into Playwright browser automation; real-time reporting captures pass/fail KPIs across 10+ test scenarios.
- • LangGraph, Google Gemini, FastAPI, Playwright, React, Python
- • Reduced manual QA scripting by 40%
IntelliLecture -Voice-to-Notes Generator
7-step async NLP pipeline integrating 5 ML models (Whisper ASR - BART summarization - Flan-T5 quiz generation - KeyBERT keyword extraction - MiniLM semantic search) to transform 1-hour audio lectures into study guides.
- • Whisper, BART, Flan-T5, KeyBERT, MiniLM, FastAPI, React
- • Async job-polling via FastAPI BackgroundTasks and Singleton lazy-loading
AI Job Application Writer -RAG Email Generator
Full-stack RAG application using LangChain + ChromaDB; scrapes job postings via BeautifulSoup, semantically matches requirements to portfolio projects, and generates tailored application emails in under 10 seconds.
- • LangChain, ChromaDB, Google Gemini, Groq Llama-3, FastAPI, React
- • Multi-provider LLM fallback delivers 99%+ generation uptime
HydrAI -Water Demand Prediction Platform
Full-stack AI product from concept to containerized deployment: multivariate feature engineering pipeline, XGBoost model achieving R² > 0.80, Gemini AI integration with multi-model fallback, CI/CD via GitHub Actions.
- • Next.js, FastAPI, XGBoost, Docker, GitHub Actions
- • In-memory caching eliminates redundant LLM calls; Docker Compose one-command deployment
Lithology Classification -ONGC
Supervised ML models (Random Forest + XGBoost) trained on 15,000+ well log records (GR, RHOB, NPHI, DEPTH) for lithology prediction; achieved 87% accuracy via systematic feature engineering, cross-validation, and hyperparameter tuning.
- • Scikit-learn, XGBoost, Random Forest, Python
- • 87% accuracy on 15,000+ geological records
KMeans Electrofacies Clustering -ONGC
Unsupervised K-Means clustering for electro-facies identification on well log data; improved reservoir zone interpretation efficiency by 25%; engineered data-driven preprocessing pipelines for large-scale geological datasets.
- • K-Means, Scikit-learn, Python
- • 25% efficiency improvement in reservoir zone interpretation
Svastya Shopify Stores (Freelance)
Built 2 production Shopify stores for Svastya Organic Farms and Svastya Wellness Store: custom Liquid sections, Dawn theme, Swiper.js sliders, product filters, mobile-optimized SEO, and third-party integrations.
- • Shopify, Liquid, HTML/CSS/JS, Swiper.js, Shopify APIs
- • Email, reviews, WhatsApp, subscriptions, analytics integrations
GitHub Repositories
Explore selected repositories from the STUDIOUS-dev GitHub account.