Full Stack Developer & Data Analyst

Subin Shrestha

I craft data narratives and build elegant solutions that make complexity feel simple.

Currently building full-stack AI features at yPilot — shipping multi-tenant SaaS infrastructure, agentic LLM workflows on LangGraph, and cost-optimized document-processing pipelines on AWS.

About

I'm drawn to the intersection of data and design — finding patterns in chaos and crafting solutions that feel intuitive. My journey began with a Computer Science degree, where building my first (admittedly rough) web app sparked something.

I've worked across the spectrum: from a fast-moving startup where I learned to ship, to established tech companies where I learned scale, and a stint at Langara's Applied Research Center where I sharpened the data side. Now I'm at yPilot , building full-stack AI features and document-processing pipelines for regulated-industry SaaS.

When not analyzing data, I'm experimenting with AI projects or exploring new ways to visualize information.

Tools & Technologies

Python R SQL TypeScript JavaScript FastAPI React Angular Ruby on Rails Tailwind CSS PostgreSQL pgvector AWS Claude (Bedrock) LangGraph vLLM PyTorch Pandas NumPy Scikit-learn Docker Git
Subin Shrestha

Selected Work

Yelp RAG System

Retrieval-Augmented Generation system built on Yelp restaurant review data featuring a comprehensive NLP pipeline with sentiment classification, BERTopic clustering, and FAISS vector search. Includes an interactive chat interface powered by locally-run Ollama models for privacy-focused, cost-efficient querying.

    Python FAISS BERTopic Ollama spaCy

Groovy

Full-stack music sharing application built with React, Django, and Flask featuring user authentication, file upload/ streaming capabilities, and an integrated RNN - based Music Generation Model that creates original piano compositions in ABC notation.

    React Django Flask Keras RNN

Telecom Customer Churn Analysis

End-to-end data analytics project analyzing telecom customer data to identify churn patterns and retention factors using Python, achieving 75% accuracy through XGBoost Model and feature engineering. Conducted comprehensive EDA with correlation analysis, implementation of PCA for dimensionality reduction and created interactive Streamlit dashboard for stakeholder presentations.

    Python XGBoost Pandas PCA Streamlit

Street Fighter Clone

Browser-based fighting game developed with vanilla JavaScript implementing real-time multiplayer functionality, character selection system with 3 unique fighters, and responsive controls for 2-player local gameplay.

    JavaScript HTML5 CSS3

Experience

Get In Touch

Currently, looking for new Software Engineering and Data Analyst opportunities in Vancouver, BC, Canada. But my inbox is always open. Whether you have a question or just want to say hi.