Shipyard
Feb 2026 — Mar 2026
A multi-tenant deployment platform that automatically builds and serves static sites on isolated subdomains — with real-time build log streaming and a 60% reduction in deployment latency via async BullMQ pipelines.
- Reduced deployment latency by 60% via async build pipeline using BullMQ and Redis with background job scheduling and distributed worker execution
- Real-time build log streaming using Redis Pub/Sub and Server-Sent Events (SSE)
- Full multi-tenant isolation — each project served on its own subdomain from Docker containers
ContractIQ
Mar 2026
AI-powered contract risk analysis — drag-and-drop legal docs, get instant clause extraction, risk scoring, and negotiation suggestions powered by Google Gemini with structured JSON outputs.
- Shared TypeScript schema layer across frontend and backend via Turborepo monorepo + Zod validation for type-safe pipelines
- Drag-and-drop document analysis dashboard with clause filtering, risk scoring and negotiation suggestions
- Google Gemini integration for structured JSON clause extraction with predictable, validated outputs
Transcribe Mate
Nov 2025
High-accuracy speech transcription service with LLM-based post-processing — 92% accuracy through inference optimization and multilingual tuning using Faster Whisper, with Gemini-powered cleanup.
- 92% transcription accuracy via inference optimization and multilingual tuning with Faster Whisper
- Gemini-powered post-processing to reduce filler noise and preserve speaker context across languages
- Reduced deployment setup time by 50% through Docker containerization of inference and cleaning services
LifeLens
Aug 2025 — Oct 2025
A personal media intelligence platform with semantic search across images, documents, and audio — powered by a RAG-based retrieval pipeline and automated AI analysis.
- Enabled semantic search across personal media collections by building a RAG-based retrieval pipeline using pgvector embeddings and FastAPI APIs.
- Automated multi-format media analysis (images, documents, audio) by implementing AI processing pipelines with Google Gemini, OCR and speech transcription.
- Reduced user wait times for media analysis by implementing asynchronous processing with Celery workers and Redis task queues.