Deterministic AI grading assistant giving teachers consistent, bias-free feedback at scale. No hallucinations, no surprises.
Agentic AI pipelines for aesthetic and content alignment serving 75M+ users. Internal tooling, modularisation, and system design.
Gamified soft-skills platform via research-backed games. Top 5 of 600+ teams at IIT Guwahati Disrupt. Led all technical development and product design.
Developed Dhvani: weakly-supervised phonemic error detection for Hindi CAPT. Published at Interspeech 2025, Rotterdam. Proficiency, enunciation, and disfluency analysis.
Built BitPy, a variable bit precision Python3 transpiler. Static type checking via AST. Custom FPGA-like datatypes with variable bit precision integers.
Published Dhvani under Prof. Siddharth (ex-Microsoft). Built demo application. Journal extension in progress.
Founded and led the AI club at Plaksha with a team of 18. Applied ML workshops, RL project teams, democratising AI education.
TA for Computational Thinking — DSA in C.
Fluid dynamics simulation using OpenDrift and PyGame.
Novel networking project under Prof. K. Gopinath.
Designed and built UI/UX for stemx.education.
“Dhvani: A Weakly-supervised Phonemic Error Detection and Personalized Feedback System for Hindi”
Arnav Rustagi, Satvik Bajpai, Nimrat Kaur
A novel deep learning approach for computer-assisted pronunciation training (CAPT) for Hindi. Weakly-supervised learning detects and localises phonemic errors with personalised user feedback — no expensive phoneme-level annotations required.
Activities: Programming, music. Research: Dhvani (Interspeech 2025).
Activities: Programming, Badminton.