Hi, I'm Avinash 👋
6th Semester Information Science and Engineering Student at RV College of Engineering
AA

About

I am an Information Science student at R.V College of Engineering. I have a strong interest in Web Development and AI/ML projects, with hands-on experience in developing practical solutions using these technologies.

Work Experience

H

Honeywell

Aug 2025 - Oct 2025
Tech Intern
  • Built an automated pipeline to convert FAA Airport diagrams into Airport Map overlays in cockpits for pilots.
  • Automatically deconstructed diagrams into layers of taxiways, runways, and labels stored in a coordinate-based format.
  • Reconstructed features from coordinates and overlaid them on satellite imagery to create airport maps with clear runway and taxiway features with labeling.
H

HPE

Feb 2025 - June 2025
Project Intern
  • Developed data collection and processing pipelines for cloud API traffic to identify SaaS applications/activities using AnyProxy, CodeBERT for labeling, and DeBERTa for unsupervised labelling.
  • Trained a RandomForest classifier (achieving 91% accuracy) for SaaS application/activity classification, with the pipeline managed and visualized via a Next.js dashboard.
  • Created a C code generation system to convert the trained RandomForest model into an optimized executable for efficient deployment on embedded systems.

Education

R

R.V College of Engineering

2023 - Present
B.E in Information Science and Engineering (ISE)

CGPA: 9.62

Skills

Python
C
TypeScript
React
Next.js
TailwindCSS
FastAPI
Electron
Postgres
My Projects

Check out my latest work

I've worked on a range of projects, from basic websites to more advanced web applications that use AI/ML. I enjoy building apps that bring together both web development and AI to create smart, useful solutions.

AI Keyboard (Tabby) - Intelligent Input Layer

AI Keyboard (Tabby) - Intelligent Input Layer

A system-wide AI assistant that transforms your input device into a real-time collaborator. Features an Interview Copilot with automated screen capture for technical analysis, an Action Menu for context-aware text transformations, and a Brain Panel for memory management using Neo4j knowledge graphs. Supports invisible typewriter mode for undetectable AI typing, voice transcription, and full Windows MCP integration for desktop automation across any application.

Electron
Next.js 15
TypeScript
TailwindCSS
FastAPI
Python
Mem0
Neo4j
Supabase
Windows MCP
Vercel AI SDK
PaperPilot

PaperPilot

An AI-powered LaTeX workspace and specialized IDE for academic writing. Features autonomous research agents integrated with Semantic Scholar and arXiv via MCP for automated literature reviews and citations. Includes a writing agent for document restructuring, a real-time Monaco editor with live PDF preview, and autonomous compilation error resolution.

Next.js 15
Tailwind CSS 4
Framer Motion
Supabase
FastAPI
Docker
TeX Live
Vercel AI SDK
Federated Learning Orchestration

Federated Learning Orchestration

A federated learning orchestration stack built on Flower (flwr) with Supabase-backed persistence. Features an admin server for coordinating SuperLink/SuperNode processes and storing run metadata, log streams, and aggregated metrics. Includes a Next.js frontend dashboard to monitor real-time training metrics and project topology.

Python
FastAPI
Flower (flwr)
Next.js
TypeScript
TailwindCSS
Supabase
OmniAgent - Multi-Agent AI System [HackRX 6.0 1st Place Winner]

OmniAgent - Multi-Agent AI System [HackRX 6.0 1st Place Winner]

OmniAgent is an advanced multi-agent AI system that autonomously processes documents, automates web tasks, executes code, and answers complex queries using LLM agents and Model Context Protocol (MCP) servers. Features include multi-format document processing (PDF, PPTX, DOCX, Excel, images), advanced OCR, web automation, code execution (Python, JS, shell), repository/code analysis, and real-time streaming AI responses. Winner of Bajaj Finserv HackRX 6.0 (1st place out of 10,000+ teams).

Next.js 14
TypeScript
Python
FastAPI
TailwindCSS
Supabase
LangChain
VoltAgent
RAG
Aviation Weather [Honeywell Design-a-thon 2025 2nd Place Winner]

Aviation Weather [Honeywell Design-a-thon 2025 2nd Place Winner]

A web application providing real-time weather briefings for flight planning, simplifying data analysis with an intuitive interface and AI summaries. Features flight planning, METAR/SIGMET/PIREP/TAF reports, AI summary, voice assistant, and interactive map. Won 2nd place at Honeywell Designathon 2025.

Next.js
TypeScript
TailwindCSS
shadcn/ui
Leaflet.js
Flask
FastAPI
Python
Gemini
Vercel
Render
Health Monitor

Health Monitor

A comprehensive health monitoring application with real-time monitoring, activity tracking, and nutrition logging. Features Arduino sensor integration for live health metrics, RAG-powered health insights using Langchain, and multimodal AI assistants powered by Gemini 2.0 and OpenAI.

Next.js
TypeScript
TailwindCSS
FastAPI
Python
Supabase
Langchain
OpenAI
Cohere
Arduino
ESP8266
Vercel
Hackathons

I like building things

Total11Hackathons
Won6Awards
Finals8Finalists
Biggest10k+Participants
  • Z

    Zenith Hackathon - AI Keyboard (Jury's Special Mention)

    Offline

    Developed an intelligent system-wide input layer that transforms any application into an AI-enabled workspace. Integrated an Interview Copilot with real-time screen analysis, context-aware autocompletions using Vercel AI SDK, and a persistent memory system utilizing Neo4j knowledge graphs and Supabase vector stores. Top 4 out of 1000+ submissions and received special jury mention.
  • M

    MindGames Arena NeurIPS 2025 - Revac (1st Place)

    Online

    Won 1st place in the Social Deduction Track (Open Division) at the MindGames Arena during NeurIPS 2025. Developed Revac_8, a sophisticated multi-module AI agent that integrates memory-based player profiling, social-graph analysis of accusations, and dynamic tone selection for adaptive communication in the game Secret Mafia.
  • B

    Bajaj Finserv HackRX 6.0 - OmniAgent (1st Place)

    Offline

    Won 1st place out of 10,000+ teams at HackRX 6.0 for developing OmniAgent, a full-stack multi-agent AI system. OmniAgent autonomously processes documents (PDF, PPTX, DOCX, Excel, images), automates web tasks, executes code (Python, JS, shell), and answers complex queries using LLM agents and Model Context Protocol (MCP) servers. Features include advanced OCR, RAG, web scraping, repository/code analysis, and real-time streaming AI responses.
  • H

    HAL Aerothon - IronEye (3rd Place)

    Offline

    Won 3rd place among 85+ teams at the HAL Aerothon for developing IronEye, an AI-powered EO/IR vision system for military object detection. The solution featured real-time RGB-IR fusion, YOLOv8 classification, and MCP-controlled War Thunder simulation integration.
  • H

    Honeywell Designathon 2025 - Aviation Weather (2nd Place)

    Offline

    Developed a web application providing real-time weather briefings for flight planning. Features included flight planning tools, METAR/SIGMET/PIREP/TAF reports, AI summaries, a voice assistant, and an interactive map. Secured 2nd place in the competition.
  • H

    HPCC Systems Hackathon - FindMySafeHaven (1st Place)

    Online

    Achieved 1st place in the hackathon with around 30 teams. Developed a platform to help travelers stay informed about the safety of their surroundings by analyzing social factors like poverty, unemployment, and education. Used the HPCC Systems platform to analyze the data.

Hobbies

Speedcubing

Solving Rubik's cubes and other twisty puzzles as quickly as possible. I've been competing since 2016 and have attended 18 official competitions.

WCA Profile

Long Distance Running

Strava

Cycling

Strava