Created ShipMind, a context-aware personal agent system with Postgres + pgvector for RAG-based memory, Bun + Fastify backend for routing and session management, and a Python-based microservice that standardizes multi-provider LLM execution; currently architecting the core agent framework

Developed Python and Rust based LLM API gateway microservices integrating OpenAI, Anthropic, xAI, and Google, standardizing model access, prompt/response formats, vector embeddings and token usage across 25+ models

Introduced Rust for concurrent processing of 12TB of data (~12m files)

Optimized CI/CD pipeline by migrating from ESLint/Prettier to Biome, reducing build times by 90-180 seconds per PR

Built and integrated multiple shipping document scrapers in NestJS (Node.js/TypeScript) within a distributed system processing 30-40k nightly tasks

Increased system reliability under unstable data sources by implementing exponential backoff, retry logic, and fault-tolerant scraping architecture

Designed PostgreSQL/Diesel ORM schema and deployed on AWS infrastructure with Terraform, Docker, and automated scheduling

Achieved 20–30% storage savings on test datasets by prototyping a Rust-based encrypted archival system with chunk-level de-duplication and asymmetric encryption via age; added Amazon S3 support and a CLI for metadata tagging, search, and selective restore

Built real-time weather monitoring firmware in Rust for RP2040 and STM32 microcontrollers interfacing with BME680 sensor via I2C, optimizing memory usage to <32KB RAM

Established code quality standards, increasing test coverage by 40% for Node.js and Go codebases

Developed Go bindings for Chia Network's BLS Signatures C++ library using CGo for cross-language cryptographic support

Engineered blockchain sync system reducing full-node sync from 3+ days to 5 minutes, enabling rapid deployment of devnets with 100+ nodes (built on AWS: Lambda, EC2, EBS)

Modernized deployment process using reproducible infra-as-code practices (Terraform, Ansible, Docker on AWS), eliminating frequent crashes and achieving 99% uptime