I'm a Full-Stack Developer allured by Machine Learning and Scalable Systems.
Frequent contributor to Open Source projects: AsyncAPI, Kubeflow, TheAccordProject.
Currently pursuing BTech CSE, pre-final year.
Looking for an intern? reach out here
C
C++
Python
Javascript
Typescript
Rust
NodeJS
BunJS
ReactJS
NextJS
ExpressJS
Bootstrap
Flask
Fast API
HF Transformers
Django
Langchain
Langgraph
SciKitLearn
PyTorch
JWT
Git
Linux
Github
Docker
Kubernetes
MongoDB
PostgreSQL
GCP
PrismaIndian Space Research Organisation (ISRO), Bangalore.
- Built an Internal library for monitoring real-time satellite telemetry data, and visualising it in spreadsheets and HTML tables.
- Mentored by Senior Scientists to build a vector knowledge base that analyses telemetry data with a React interface, shipping an end-to-end RAG system.
Stealth Startup, Remote.
- Shipping AI pipelines for data sensitive documents extraction and web interfaces and real time communication to the backend to show live extraction progress.
- Tinkered with multiple SOTA OCR models to optimize text and table extraction, while building production ready APIs and maintaining deployments in a VPS.
Onebit AI, Remote.
- Architected a React Native app to harness Llama.cpp, Bitnet.cpp and Whisper.cpp to be a meeting assistant that listens and summarizes, working fully on CPU inference.
- Analysing full-precision model performance post QAT/PQT as part of bringing frontier model capabilities to local 1.58bit models.
Glyph
Glyph is an agentic chat platform, with production ready RAG, file ingestion pipelines. Including custom Manim Diffusion infused illustrations to enhance product launches and illustrate complex systems.
Problem
- Very niche product launches lack depth and detail with diffusion based video generation
- In specific industries where products demand detailed analyses, current tools fail to depict complex visuals easily
Solution
- Glyph uses Manim, an open source library, to create explanatory visuals for product launches.
- Glyph introduces State-of-the-art Agentic Orchestration which enhances the end quality of the video by splitting it into multiple scenes and stitching them together.
Nemo
Nemo is an AI productivity app focussing on Edge AI and Local inference of SLM's, and optimising its performance suitable for daily use and finance tracking, opening multiple avenues for building features.
Problem
- Cloud-based AI assistants pose privacy risks by storing sensitive personal data on external servers.
- Latency from remote AI inference disrupts real-time productivity workflows and conversational flow.
Solution
- Nemo runs locally on-device using Llama.cpp for private, secure AI interactions without sending data to the cloud.
- Edge AI with SLMs enables instant responses for real-time productivity and seamless conversations.
Lippy-Agent
Lippy brings you leads from any website changes — an Email, Slack, or Telegram DM about the changes. Monitor your competitors, check price and place an order for your Mac. Do anything, over an MCP server!
Problem
- Competitors change pricing and content silently, and manual monitoring means catching updates late.
- Website-change alerts are scattered across emails and feeds with no unified, actionable channel.
Solution
- Lippy watches target sites and pushes Email, Slack, or Telegram DM alerts the moment anything changes.
- Scheduled diffs against competitors surface only the meaningful updates, killing the noise.
Voice Diarisation Pipeline
A local-first pipeline to diarise any audio file, supporting multiple codecs thanks to FFMPEG. Includes an identification pipeline to identify the speakers and classify male and female voices.
Problem
- Most diarisation services upload sensitive audio to the cloud, raising privacy concerns.
- Narrow codec support forces manual transcoding before any processing can start.
Solution
- Runs fully local-first — no audio ever leaves the machine.
- FFMPEG integration ingests any codec out of the box, no manual transcoding.
CMR University, Bangalore
