I build backend, ML, and AI things, and I like it when they actually ship.
Software Engineering Intern · StellantisVirtualization and Python/QEMU automation.
Backend Developer Intern · OTET InfosystemsLaravel to Django migration and REST APIs with DRF.
Live Company CheckLive AI company check with a human reviewer in the loop.
BehaveGuardContinuous biometric auth from keystrokes and mouse.
CortexRAG companion that answers questions about a codebase.
MAIT DelhiB.Tech IT · 9.30 CGPA · 2027
IIT MadrasBS Data Science · 8.93 CGPA · 2027
I'm a software engineering intern at Stellantis and a student at MAIT Delhi and IIT Madras. Most of what I make lives somewhere between a Python API, a queue, and a model that needs a home.
Outside of work you'll find me tinkering with side projects, tuning my dotfiles, or reading manga. The background of this page is Conway's Game of Life. Go ahead, poke it.
Just started. Working on software virtualization with QEMU, and automating the workflows around it with Python.
OTET Infosystems
Helped migrate a Laravel app to Django, built and updated REST endpoints with Django REST Framework, and hunted down bugs across models, serializers, and API responses.
An AI agent that checks a company's records, customer reviews and news while a reviewer watches live, answers its questions, pauses it and sends feedback. Charts are computed from the data, never written by the model.
The agent only describes what to plot; code computes every number, so charts and findings stay traceable to the data.
Continuous biometric authentication from how you type and move your mouse. BiLSTM + TCN fusion, async retraining over Redis Streams.
Fast classical scoring answers in milliseconds, while the neural models retrain in the background and only get promoted after beating a held-out validation split.
A RAG codebase companion. Indexes repos and pull requests with AST-aware chunking, then answers questions with cited code over streaming chat.
Splits code along function and class boundaries with tree-sitter instead of fixed token windows, so syntax and scope survive indexing.
Sub-5ms terminal autocorrect in Go. Uses LLM context and directory awareness, with a safety tier that stops it from suggesting destructive commands.
Every suggestion is classified as safe, warning or danger, and destructive commands need an explicit yes.
Plant leaf super-resolution with an ESRGAN-lite conditional GAN, RRDB blocks, EMA checkpoints, and CI smoke tests.
Mixed-precision training and EMA checkpoints keep the GAN stable and its outputs free of oscillation artifacts.
Four-class toxicity classifier with 66+ engineered features, out-of-fold target encoding, blend optimization, and tuned per-class thresholds.
Self-hosted secrets vault with PBKDF2 key derivation and Redis-backed, session-only key handling.
Master keys are derived with PBKDF2 and live only in an in-memory Redis session, never on disk.
LLM agent that browses, extracts, runs sandboxed code, and answers quiz questions over an API.
B.Tech in Information Technology · CGPA 9.30
BS in Data Science and Applications · CGPA 8.93
I'm always up for talking about backend systems, ML, or whatever you're building. My inbox is open.