I build systems that look at things and work out what they are. Currently a research intern at IIIT Allahabad, teaching a network to read currency from a phone photo - and another to notice when a student is struggling.
Two systems, both built to notice something a person would otherwise miss.
A convolutional network that identifies Indian currency denominations from an ordinary phone picture - the kind taken one-handed, badly lit, at an angle. Built during a research internship at IIIT Allahabad under Dr. Shiv Ram Dubey, with accessibility for visually impaired users as the point of the exercise.
Institutions lose track of their alumni the moment they graduate - manual surveys go stale and nobody updates them. This tracks career movement automatically: it scrapes public professional profiles, detects when a company, designation, location or role changes, and keeps the old record instead of overwriting it, so the whole career trajectory survives.
Paper Agrawal, Sutaone, Kumari, Tiwari & Ghosh - Automated Alumni Career Tracking using Web Scraping, Change Detection and Data Analytics
I'm a Computer Science undergraduate at IEM Kolkata, currently third semester and holding a 9.43 - which sounds better than it feels at 2am the night before a viva.
Most of my work sits in computer vision: taking something visual and messy and getting a machine to be reliably right about it. I like the part where the accuracy is bad and you have to work out why. That's the actual job.
Outside that - I've presented at SPECTRUM (and won best poster), represented my college at NUS Singapore, and sat through IIT Guwahati's immersion programme taking far too many notes.
Open to AI/ML engineering and software roles, research collaborations, or an argument about whether accuracy is a useful metric.
s.ghosh260972@gmail.com