I build high-performance machine learning systems, agentic AI products, and full-stack tools. I specialize in designing explainable predictive models, deploying tree-based ensembles, and orchestrating multi-agent systems.
Machine learning system to analyze stock data and optimize portfolio allocation. Applied data preprocessing, feature engineering, and risk-return modeling.
- Live Demo: red-glacier-gamma.vercel.app
- Tech Stack: Python, Scikit-learn, Portfolio Theory, Monte Carlo simulation, Gemini API sentiment analysis
Explainable credit-scoring underwriting engine for India's 190M+ underbanked population using gradient-boosted ensembles.
- Metrics: 0.9134 AUC-ROC and 0.6910 KS statistic on alternative UPI & utility payment signals.
- Live Demo: kreditflow.vercel.app
- Tech Stack: XGBoost, LightGBM, CatBoost, Scikit-learn, FastAPI, Next.js, TypeScript, Tailwind CSS, SHAP Explainability
Multi-agent business planning platform using a sequential pipeline of virtual executive agents to audit, model, and roadmap startup ideas.
- Live Demo: aethercoo.vercel.app
- Tech Stack: Agentic AI, Gemini API, Next.js, React, JavaScript, Vercel
- Machine Learning & Stats: XGBoost, LightGBM, CatBoost, Ensemble Methods, Monte Carlo Simulation, Risk-Return Modeling, SHAP Explainability, Scikit-learn, Feature Engineering
- Generative AI: Agentic / Multi-Agent Systems, NLP & Sentiment Analysis (Gemini API)
- Development & Frameworks: Python (NumPy, Pandas, Scikit-learn), SQL, Git & GitHub, Next.js, FastAPI, Node.js, Vercel
- Design & IoT: UI/UX Design (Figma), IoT & Real-Time Sensor Data Processing
- Portfolio: utkarsh-shankar-portfolio.vercel.app
- LinkedIn: linkedin.com/in/utkarsh-shankar12
- Email: utkarshshankar1456@gmail.com