CS @ Umass Amherst | AI/ML • Reinforcement Learning • Quant Finance • Computational Neuroscience
- 🎓 CS Student @ UC Riverside (GPA: 3.97) — IB Diploma, Indus International School Pune
- 🔬 Undergraduate Researcher — Hyperspectral Imaging with Prof. Saragadam (tensor decomposition: SVD, CP, Tucker vs. neural implicit modeling, 98.7% reconstruction accuracy, targeting IEEE CVPR)
- 📡 Co-author on exoplanet habitability research — Bayesian hierarchical modeling + Random Forest on 4,510 NASA Exoplanet Archive planets
- 💊 Co-author on AI in Drug Discovery & Materials Science — Springer Nature submission (41-page review, mathematical frameworks + GNNs + topological data analysis)
- 📐 Co-author on Mathematics & Physics in the 21st Century — AI-augmented discovery, Langlands program, TQFT, ECC
- 📦 Published IEEE-format paper on ARIMA Inventory Forecasting — RMSE 15.05, Walmart benchmark validation
- 🤖 Author — Forecast-Seeded Non-Stationary RL (FS-NRLF): 16–21% profit gain, 54.8% stock-out reduction over PPO on M5 benchmark (3,049 products, 5.4M records); under double-blind review
- 🏭 Author — Modern Digital Supply Chain Architecture white paper: integrated AI forecasting (ARIMA/LSTM/Transformer), IoT sensing, blockchain traceability, RPA automation, and SAP S4HANA ERP
- 🏆 Silver Medal — 5th International STEM Olympiad (Top 5 globally, Frankfurt); Top 200 Under-19 Chess (AICF); 2× District Champion
- 🎯 Targeting FAANG SWE/MLE internships and Quant/Research Engineer roles
| Project | Description | Stack |
|---|---|---|
| Habitability Assessment | 3-model framework (Bayesian MCMC + Random Forest + Logistic Regression) to correct transit survey bias on 4,510 exoplanets. R²=0.835, 100% HZ recall, 36 PLATO candidates | Python, emcee, scikit-learn, NASA API |
| Inventory Quantum | Quantum-Enhanced Retail Optimization — 22,500-line system integrating Temporal Fusion Transformers, QAOA quantum algorithms, Conservative Q-Learning. NeurIPS-format paper authored | Python, FastAPI, Docker, Qiskit |
| NeuroForge | Biophysical neuron simulation engine with GPU acceleration | Python, Julia, CUDA |
| RL Portfolio Rebalancing | Deep RL agent for dynamic portfolio optimization | Python, PyTorch, FinRL |
| ARIMA Inventory Forecasting | Multi-dimensional ARIMA evaluation (accuracy, complexity, robustness, stability) — RMSE 15.05 on primary retail data, Walmart secondary validation | Python, statsmodels |
| AutoApply | Autonomous job application platform with AI-generated CVs | Next.js, TypeScript, AWS S3 |
| ML Platform | 15,000-line ML platform — Custom Neural Bradley-Terry architecture, 0.987 AUC-ROC, Monte Carlo simulation (10K iterations), AWS EC2 + PostgreSQL | Python, AWS, PostgreSQL |
| Paper | Venue | Status |
|---|---|---|
| Forecasts as Priors over Initial States: Sample-Efficient RL for Predictable Non-Stationarity (FS-NRLF) — 16–21% profit gain, 54.8% stock-out reduction over PPO on M5 (3,049 products, 5.4M records) | Under Double-Blind Review | Submitted |
| Modern Digital Supply Chain Architecture — Integrated AI forecasting (ARIMA/LSTM/Transformer ensemble), IoT sensing, blockchain traceability, RPA automation, SAP S4HANA ERP | White Paper | Published |
| Mathematical Frameworks and AI Applications in Drug Discovery and Materials Science | Springer Nature | Under Review |
| Mathematics and Physics in the Early 21st Century: Foundations, Collaboration, and AI-Augmented Discovery | Springer Nature AI | Under Review |
| Integrating ARIMA, Quantum Decision Models, and RL for Omnichannel Retail | Springer Nature | Under Review |
| Multi-Dimensional ARIMA Performance Analysis for Inventory Forecasting | IEEE | Under Review |
| Quantifying Detection Bias and Recovering Habitable Zone Occurrence Rates (3-Model Framework) | Conference-targeted | In Preparation |
AI/ML Engineering Intern — SAP (Dec 2024 – Present, Part-time)
- Implementing production AI for enterprise workforce management serving 10,000+ users
- Conducting A/B testing and deploying ML models (Python, TensorFlow)
AI Research Specialist — AWS & NUS Singapore (Dec 2023 – Jan 2024)
- Competitive international program; achieved 99.8% model accuracy — highest in cohort (12 countries)
- Developed transfer learning + model compression for edge deployment
Undergraduate Researcher — UC Riverside (Prof. Saragadam) (Sep 2025 – Present)
- Hyperspectral imaging: comparing tensor decomposition (SVD, CP, Tucker) vs. neural implicit modeling
- 98.7% reconstruction accuracy; targeting IEEE CVPR publication
Founder & Program Director — BRIDGE Interdisciplinary Research Initiative @ UCR (Oct 2025 – Present)
- 15 STEM + 5 humanities/social science underclassmen; 8 collaborative research projects
- 2 conference presentations; 2 department partners
Founder — AI for Community Impact Program (Jan 2024 – Present)
- Connected 15 HS students with 8 nonprofits for ML solutions; 30% efficiency improvement for 3 orgs
- Mentored 12 students in Python/DS
Curriculum Manager — Daricha Foundation (May 2024 – Feb 2025)
- Led curriculum for 200 underprivileged students across 5 centers
- Raised course completion from 45% → 85%; trained 15 instructors; 35% test score improvement
- 🥈 Silver Medal — 5th International STEM Olympiad (Top 5 globally, 500 students, 40 countries, Frankfurt)
- ♟️ Top 200 Under-19 India Chess — AICF; 2× State representative at Nationals; 2× District Champion
- 🏓 State Table Tennis — Divisional & District Champion (TTFI)
- 🥇 1st Place — Code Day Hackathon (gesture recognition, 92% accuracy, OpenCV + CNNs)
- 🏅 Top Performer — AWS & NUS AI Research Program (99.8% accuracy, highest in cohort)
- 📜 IBM Full Stack Development Professional Certificate — React, Node.js, Docker, Kubernetes, CI/CD
Languages: Python · C++ · Julia · TypeScript · SQL
ML/AI: PyTorch · TensorFlow · scikit-learn · Hugging Face · spaCy · emcee (MCMC)
Quant/Finance: FinRL · QAOA · QUBO · Conformal Prediction · ARIMA · TFT
Research: Tensor Decomposition (SVD, CP, Tucker) · GNNs · Topological Data Analysis · Bayesian Inference
Infra: AWS · Docker · FastAPI · Next.js · PostgreSQL · Redis
Tools: Git · Linux · Jupyter · LaTeX
"Build things that are hard to build."