Analyst, Operations Performance & Analytics @ United Airlines · Data Science · Machine Learning · Building data products
Working toward a Master's in Data Science with a focus on AI/ML
- I build analytics, forecasting and decision-support tools at United Airlines. My work there has helped unlock $77M+ in operational impact.
- I build predictive models (regression, time series), ETL pipelines in Python / SQL / PySpark, and LLM-powered self-service analytics.
- Awards: Sky Laureates Award (2026), IKC Ambassador Award (2026), 1K Category Award (2025) at United Airlines.
- Education: B.Tech, Electrical Engineering, Punjab Engineering College (CGPA 8.2).
- Next: a Master's in Data Science, going deeper into machine learning, deep learning and applied AI.
Data science & ML
Analytics & BI
Building products
| Project | What it does |
|---|---|
| GoBizSchool | A business-school discovery and application platform. It tracks deadlines, rounds and school profiles for top MBA programs and helps you plan each application through to submission. |
| Portfolio | My personal site: experience, skills, projects and awards. |
| Resume Studio | Write a LaTeX resume in the browser with a live preview and a print-ready PDF. No sign-up, no server. |
| Referrals | A structured way to request a United Airlines referral from me, so every request arrives with the details I need. |
| Tools by Yash | Private developer utilities that run entirely in the browser: JSON, cron, tokens, encoding, colors and more. |
| Links | One map of everything I'm building. |
A collection of Python projects spanning simulation, AI reasoning and reliability, data quality, execution tracing, and static analysis. They make systems easier to model, inspect, and evaluate with reproducible evidence.
| Repository | What it does |
|---|---|
| SimulSI | An experiment-first discrete-event simulation framework for operational systems. Model queues, resources, and disruptions, then compare scenarios with reproducible seeded runs, paired confidence intervals, Monte Carlo, and sensitivity analysis to turn “what if?” questions into evidence for decisions. |
| AdverSI | Adversarial, evidence-aware multi-agent reasoning. Agents reason independently, make typed claims, cite sources that are checked word for word, cross-examine each other, and hand a judge a record whose verdicts are validated by deterministic rules. |
| SynapSI | Collective AI reasoning. Multiple agents exchange structured claims and evidence, make disagreement explicit and synthesize a result, with an experiment engine that measures when more agents help and when they hurt. |
| RedSI | Red-teaming and reliability engineering for AI systems: discover, reproduce, measure and prevent failures, with regression gates for CI. Work in progress. |
| DataSI | Data quality investigation. Finds missingness patterns, duplicates, type problems, outliers, train/test drift and features that predict the target suspiciously well, using statistics rather than LLMs, and separates what was measured from what it might mean. |
| AuctionSI | Marketplaces and auctions for autonomous agents. Agents bid on tasks, a pluggable mechanism picks an explainable winner, and the work is contracted, verified deterministically, settled and fed into multi-dimensional reputation, with simulations to compare market designs. |
| ReconSI | Data reconciliation. Finds out why two datasets that should agree don't: missing and extra records, duplicate and malformed keys, value differences under explicit tolerances, grain mismatches, systematic bias, where mismatches concentrate and when they started, with every finding labelled as observed, statistical or a likely explanation. |
| PyInsider | A local-first Python execution recorder that captures function calls, timings, arguments, return values, and exception paths, then lets you explore the recorded call tree in an interactive viewer. |
| project-mapper | A static Python dependency mapper that recursively traces local imports from an entry point and generates a self-contained HTML report, without executing the code. |
| claude-session-handoff | A Claude Code skill that keeps local project checkpoints: decisions, edits, verification, and next steps, so a new session can resume after a reset or lost context. |