Aditya Shukla

I worked on the team that collects retail data from marketplaces like Amazon, Walmart and TikTok Shop for client brands. Over the summer I owned 10 live Airflow and Spark pipelines and shipped 19 production changes across 3 codebases. The highlight was tracking down a silent bug, caused by expired cookies, that had cut our Amazon product coverage from 11K a week to 3K, then fixing it. I also found a 1,668-order gap that had gone unnoticed since February, built a new Amazon deals data source from scratch, and turned a repetitive backfill task into a reusable tool the team now uses.

Netradyne had useful internal AI tools, but only developers could reach them, since they ran inside coding environments (IDEs). I built FRIDAY, a chat platform that let anyone on the business side type a request in plain English and have the right tool run for them. I built it end to end: a React and TypeScript frontend, a FastAPI backend with MongoDB for chat history, and Azure OpenAI to work out which tool to call and with what inputs. I also packaged it in Docker so several tool servers could run behind one interface.

I spent six weeks in Japan working on next-generation air-quality sensors, mostly on my own in a lab I had never worked in. I handled the whole build: growing ultra-thin indium oxide (In₂O₃) nanosheet films using a liquid-metal method, adjusting the processing temperature by alloying with bismuth, and turning the films into working gas-sensing devices on Cr/Pt electrodes. I set up automated resistance logging and analysed the time-series data, which showed a clear 303% response to nitrogen dioxide (NO₂) above a noisy baseline and confirmed the devices work. Learning to work independently in a new country and a new lab was as valuable to me as the research itself.