Helping rail teams spot problems earlier
TrackSense AI
I am building TrackSense with my co-founders: edge sensors and local models that flag possible track problems while keeping the final decision with the operator. My CV records a paid Indian Railways pilot for structural monitoring on the Yamuna crossing in Delhi.
TrackSense · Try the demo
Human-in-the-loop by design · operator decides
Watch a small change become an alert.
This is a simple simulation, not live railway data. Five trackside sensors show vibration, sound, and heat. One begins to behave differently, the model flags it, and then the demo waits for you to decide what happens next.
TrackSense · System path
From trackside signals to an accountable decision
Edge sensor nodes
Vibration, acoustic, and thermal signals are gathered close to the track.
Local anomaly filter
A local model ranks unusual changes and carries uncertainty forward.
Operator command view
Evidence is presented for review; the responsible person makes the call.
Designed to keep operational decisions with the human responsible—not to turn a model score into an automatic shutdown.
I made this with generated sample data to explain the idea. It is not field data and it does not claim production performance.
“I do not assume systems are infallible. I design for the moment technology is wrong.”
I built the pause into the demo on purpose. The model can point to a problem, but the person responsible for the railway must still make the call.
The problem
Rail inspections often happen in separate, periodic steps. By the time signals are brought together, maintenance teams may have lost valuable time.
My part in the work
As co-founder and technical lead, I turn what operators need into the product architecture, workflow, and technical plan.
What was involved
What I did
- I organised the product around three plain questions: what do we see, what might it mean, and what should the operator do next?
- I designed a concept that brings video, vibration, thermal, and acoustic signals together.
- I made confidence, uncertainty, and human review part of the product rather than an afterthought.
- I connected the product requirements to a realistic pilot plan, costs, and deployment limits.
What changed
- Best Overall Project — IE School of Science & Technology Venture Bootcamp.
- Ranked first among 75+ ventures at IE; Best Pitch at IE Startup Lab.
- A paid Indian Railways pilot for structural monitoring on the Yamuna crossing in Delhi, as listed in my CV.
- A working interface that lets operators review signals and make the call.
Why I include this work
I include TrackSense because it brings together my engineering background, AI product work, and belief that people must remain accountable for consequential decisions.
See the live project