I build robust, scalable software and intelligent systems — from low-level C++ performance optimization to multi-agent AI orchestration platforms.
What I work with:
- Languages: Python, Java, C++, TypeScript, JavaScript, SQL, Kotlin
- AI/ML: Multi-Agent Orchestration, LLM Integration (Claude, OpenAI), Vector Databases (pgvector), Reinforcement Learning, ONNX Embeddings
- Backend: FastAPI, Node.js, Celery, Redis Streams, WebSockets, REST APIs
- Frontend: Next.js, React, Tailwind CSS, DuckDB-Wasm
- Mobile: Android, Kotlin, Jetpack Compose, Bluetooth/Wi-Fi Direct, P2P Mesh
- Infra: Docker, PostgreSQL, Linux VPS, Grafana, OpenTelemetry, Caddy
Highlighted Projects:
- Amber — Personal memory vault with hybrid search (BM25 + vector embeddings via Reciprocal Rank Fusion), CLI, web omnibox, and browser extension.
- Job Harness — Agentic job-application pipeline sourcing from 5 ATS APIs, LLM-scored fit ranking, tailored resumes, and human-in-the-loop browser automation.
- Aegis — Multi-agent automation platform with Celery, Redis Streams, OpenTelemetry, and HITL policy engines.
- Zik — Offline-first, zero-ads, open-source Android music player. Kotlin, Compose, Media3, gapless playback, folder browsing, parametric EQ.
- Nexa — P2P offline-first messaging with DTN routing, Bluetooth/Wi-Fi Direct mesh, and Jetpack Compose UI. UCT Capstone project.
- FRETAK Protocol — 9-container microservice fleet with AI translation (Meta NLLB) on bare-metal Linux VPS.
- SmartDocs — Multi-format RAG pipeline with hybrid semantic/keyword search and citation-backed responses.
- AgentSwarm — Multi-agent orchestration framework with specialized agents and Redis Streams messaging.
- Posthoc Terminal — In-browser analytical environment querying millions of records at 60 FPS via DuckDB-Wasm.
Awards:
- 1st Place — Entelect University Cup II (2025)
- 2nd Place — Predictive Insights Hackathon (2025)
Currently seeking: Remote AI Automation / Agentic Python / Software Engineering roles — flexible schedules preferred.