Explore My Creations
Projects
Realis
Verifies API docs against live API behavior and auto-generates typed SDKs from verified truth
- - Built a platform that verifies API docs against live API behavior and auto-generates typed SDKs in 3 languages from verified truth.
- - Architected a two-plane distributed system over PostgreSQL + Redis Streams with at-least-once, idempotent job processing.
- - Cut false findings via majority-vote LLM cross-verification & replayable evidence; validated by 250+ tests.
Java Spring Boot Python PostgreSQL Docker
Zeruel
Fully on-device Android voice assistant that plans with a local LLM and drives third-party apps with zero cloud dependency
- - Built a fully on-device Android voice assistant that plans with a local LLM and drives third-party apps with zero cloud dependency.
- - Engineered an 11-module Gradle architecture with faked device interfaces, enabling 177 JVM unit tests.
- - Implemented on-device semantic search (ONNX embeddings) & an MCP server so AI agents can drive the phone.
Kotlin Java
Mikuji
A proactive MCP server that pushes context to AI agents via generic watchers and a deterministic YAML policy engine
- - Built a proactive MCP server that pushes context to AI agents via 7 generic watchers and a deterministic YAML policy engine.
- - Guaranteed zero-loss delivery with a capability-negotiated fallback chain across 4 MCP primitives.
- - Logged every action in a tamper-evident hash-chained audit log, gated behind a 4-level trust ladder.
TypeScript Node.js Next.js
Zanzo
Your Instagram saves, organized and searchable — transcribed, categorized, and extracted into structured data
- - Built a self-hosted Instagram saves organizer with automatic ingestion via a poller and DM bot, requiring zero manual URL pasting.
- - Engineered a category-aware AI pipeline (Whisper + Gemini) that extracts structured schema per category — recipes yield ingredients & steps, events yield dates & venues.
- - Implemented hybrid semantic + full-text search over PostgreSQL + pgvector with multi-user JWT auth and an admin panel.
Python FastAPI Next.js PostgreSQL Docker
Documentation Assistant
MCP for Spring Boot
- - Scraped and preprocessed official Spring Boot docs, then vectorized and stored embeddings in a Vector DB.
- - Built a Spring Boot MCP server that interfaces with the Vector DB to serve relevant documentation chunks for downstream LLMs.
- - Implemented a clear separation between data collection and retrieval by isolating the scraper and MCP server components.
- - Designed for seamless integration into AI agents, developer chatbots, or knowledge assistants.
Spring Boot MongoDB
