I build AI systems that survive contact with production.
AI engineer at O5 Apparel, where I help the team automate their tech operations. Before this: data science at PayPal, satellite imagery analysis at ISRO, open-source geospatial tooling on GRASS GIS at NC State's Center for Geospatial Analytics, and computer vision work at FX is AI. The stack has changed every time; the job — turn a messy problem into something that ships — hasn't.
I'm an M.S. Computer Science graduate of NC State, by way of a B.Tech at Nirma University in India. My first real engineering job was at ISRO, India's space agency, writing machine learning models for satellite data and shipping them in Docker containers into mission-critical pipelines. From there: a summer at PayPal building analytics dashboards at scale, which earned me a Bravo Award; a semester at FX is AI building computer vision models with a frontend on top; and two semesters at NC State's Center for Geospatial Analytics, contributing modules to GRASS GIS, open-source software researchers around the world actually run. Since June 2026 I've been an AI engineer at O5 Apparel, automating the parts of their tech operations that used to need a person.
Day to day I reach for Python, PyTorch, and SQL for the modeling; React and Node when the model needs a face; Docker and GCP/Kubernetes when it needs to survive traffic. None of that is the point — the point is closing the gap between a model that scores well offline and one that holds up in production, which is where most of the interesting bugs live.
Off the clock I'm usually on a squash or pickleball court, cricket when I can find a game, tracking down good food, or building a small games hub just because school-notebook game-of-book-cricket deserves a digital version. I've come to think the stack doesn't matter much — the problem does.
Forecasting pipeline on multivariate IoT sensor data, tuned to an R² of 0.96 on held-out test data. Anomaly-detection experiments and visualizations sit on top, built to catch a sensor drifting before it triggers a false alert.
Centralized tracking for LLM prompt inputs, outputs, latency, and cost across applications. Prompt versioning and regression detection catch output drift before it reaches a user, not after.
A financial digital twin on GKE, built on top of Google's Bank of Anthos microservices, simulating affordability scenarios in real time. A Gemini-powered multi-agent chatbot serves the analysis through a live production API.
A virtual wardrobe app pairing garment detection with pose estimation, so a user can see an outfit on themselves before committing to it. Shipped as a live, public demo.
"I have had the pleasure of working alongside Jayneel during his internship and he is very dedicated. He is a quick learner and delivered the assigned technical projects successfully, taking end-to-end responsibility. Jayneel will be a valuable addition to any data analytics team."
— Samujjal Seal Sarma, Senior Manager, Data Science @ PayPal (managed Jayneel directly) · view on LinkedIn →
I'm always glad to talk shop, trade project notes, or hear about a role in AI engineering, ML, or data. Email is the fastest way to reach me.
jayneel.shah18@gmail.com