Skip to main content

Selected work.(10)

(01) Featured · Open model

FluentWhisper

Apache 2.0 LoRA adapter on Whisper large v3 turbo that removes fillers, discourse markers, repetitions, and self-repairs in one ASR pass.

WER · DisfluencySpeech test · N 250

Whisper, vanilla9.4%
FluentWhisper3.4%
02 · Voice AIFastAPI · Pipecat

Grand Plaza

Voice AI hotel concierge with room service ordering over Daily WebRTC.

03 · Voice AINext.js · Cloudflare Workers

Career Scout

Voice assistant for job hunting. Ask it to research a company, prep you for an interview, or help with outreach.

Top 3HackLLM 2025 · IIIT Delhi · 80+ teams
04 · ML + AgentsPython · Transformers

MedSum AI

Medical summarizer with dual-perspective output, built without external APIs. 3rd of 80+ teams at HackLLM 2025, IIIT Delhi.

05 · Voice AIPython · WhatsApp Business

WhatsApp Agent

WhatsApp Business voice call assistant with a Python webhook server.

06 · ML + AgentsTypeScript · AI

Mira

Upload a CSV and Mira returns insights and interactive charts built from its columns.

07 · ML + AgentsPython · LLMs

Idea Forge

An agent that brainstorms hackathon ideas with you, one round at a time.

08 · ML + AgentsJavaScript · Chrome Extension

GenEye

Chrome extension that classifies posts on X in real time with Llama 3.3 70B, for studying what a feed is made of.

09 · Web + ToolsAstro · CSS

Autokinetic

Open-source browser studio for kinetic typography. Design text animations visually and export zero-dependency CSS.

10 · Web + ToolsJavaScript · Apple Vision

Native Vision OCR

Drag a box over any part of the screen and the text lands on the clipboard. Runs on Apple's on-device Vision framework.