Open to opportunities

Sreyash S

@sreyashs

Message

I build data, machine learning, and full-stack systems for real-world decision-making.

India

Message

What I'm looking for

I'm looking to build data, machine learning, and full-stack products where I can apply quantitative modeling, robust data pipelines, and practical deployment to solve real-world problems.

I'm building data and machine learning systems that turn complex information into useful insights, from market-volatility forecasting to healthcare and agricultural applications.

My research on task-free continual learning for market volatility introduces an online Elastic Weight Consolidation approach using a Wasserstein-2 drift signal. Across a 30-seed NIFTY-50 evaluation, it reduced relative pre-crisis forgetting from 12.67 to 0.60 and variance by roughly 25×.

I've also shipped a live digital prescription platform with React, FastAPI, PostgreSQL, Supabase, and JWT authentication, and built ML solutions for brain-tumor detection, crop recommendation, yield prediction, and stock-return forecasting.

Experience

Work history, roles, and key accomplishments

Current

Full-Stack Developer

Current

Digital Prescription & Patient Management Platform

Jul 2026 - Present (2 months)

Built and shipped a live full-stack e-medicine platform using React, Tailwind, FastAPI, and PostgreSQL/Supabase, enabling doctors to generate digital prescriptions delivered to patients by email. Designed a normalized PostgreSQL schema and secured FastAPI endpoints with JWT authentication.

React Tailwind fastAPI PostgreSQL Supabase JWT SQL REST APIs

ML Developer

Calendar-Aware Stock Return Prediction Engine

Apr 2026 - Present (5 months)

Engineered a feature-engineering pipeline over 120,000+ daily market observations and developed a Regime-Gated Multi-Scale Temporal Convolutional Network (RG-TCN) with parallel multi-kernel convolutions and temporal attention. Validated multi-horizon forecasting using RMSE and Pearson correlation.

Python Feature Engineering Attention Mechanisms Pearson Correlation

Backend / ML Developer

AI-Based Agricultural Advisory System

Apr 2026 - Present (5 months)

Built a scikit-learn pipeline over 17,600+ data points with outlier capping and Yeo-Johnson transforms, feeding XGBoost and Gradient Boosting models. Tuned models via Optuna to over 90% accuracy and deployed them behind a FastAPI inference engine with 16+ REST endpoints.

scikit learn XGBoost Gradient Boosting Optuna SHAP fastAPI REST APIs Python