Real projects, real impact

AI video editors, privacy-first mobile apps, on-device ML, and full-stack platforms, built by a staff-level engineer with 12+ years of experience.

Featured

AI-Powered Video Editor

Client: ngramLead Engineer

Challenge

The existing market lacked a video editor where users could direct edits through natural language. Competing tools had no AI-native editing workflow.

Solution

Architected an AI companion that turns plain-English instructions into video edits via MCP-based tool orchestration. Built the full web editor: timeline UX, rendering pipeline, real-time state management, and collaborative editing. Shipped as lead engineer.

ReactTypeScriptAI/MLMCP Protocol
Live

Privacy-First Vehicle Platform

Client: Hello Tag

Challenge

Vehicle owners had no way to receive anonymous messages from passersby (e.g., 'your lights are on') without exposing personal contact info.

Solution

Built a QR code-based mobile app: scan a tag, send a message, and the owner gets a push notification with real-time chat, all without sharing phone numbers. Shipped to App Store in under 3 weeks.

React NativeExpoSupabaseTypeScript

On-Device AI Financial Kit

Challenge

Users needed financial insights from their email data but refused to send sensitive information to cloud APIs.

Solution

Built a fully offline Android app running TinyLlama via ExecuTorch and Qualcomm QNN. Extracts transactions from emails, stores in local SQLite, and answers natural-language queries, all with zero network calls.

Top 6 at GitHub HQ Hackathon (Meta & Cerebral Valley)

ExecuTorchTinyLlamaAndroidQualcomm QNN

AI Resume Builder

Client: CraftMyResume AI

Challenge

Job seekers spent hours formatting resumes manually. Existing AI tools produced generic, poorly formatted output.

Solution

A conversational AI that pulls data from GitHub and LinkedIn, generates tailored resume content, matches reference layouts through chat, and exports pixel-perfect PDFs.

Next.jsAIPDF Export

Intelligent Email Analytics

Client: Inbox AI

Challenge

Years of email history contain valuable financial and personal data, but manually searching through thousands of messages is impractical.

Solution

Upload a Gmail export and ask natural-language questions like 'how much did I spend on flights in 2024?' Auto-categorization, spending trend analysis, and structured data extraction powered by Claude and GPT-4.

PythonClaude/GPT-4ReactNLP

Your project could be next.

Tell us what you need built. Free discovery call, proposal within 48 hours.