Software Engineer & AI Expert

Recall is a decentralized intelligence network for AI agents (1.4M users, 175K agents evaluated). I led major initiatives: built the erasure coding and recovery system from scratch in Rust — handling async upload/recovery across IPFS and Iroh with correctness under failure as the defining requirement. Led the data availability layer integrating IPC for scaling, Filecoin for on-chain storage, and bridging with Ethereum and CometBFT. Designed the AI-blockchain sync service with production robustness and diagnosability as first-class requirements. Also worked at the protocol layer: smart contracts in Rust and Solidity, cross-chain token bridging with Axelar.

DefraDB runs directly on edge devices — IoT sensors, satellites, factory controllers — letting them manage and sync data locally without cloud dependency. I implemented core database features: secondary indexes, document encryption, digital signatures, searchable encryption (query encrypted data without decrypting), and a KMS using ECIES over libp2p. The CRDT-based sync enables full offline operation with automatic conflict-free state merging on reconnect. Used in AI edge training, satellite data management, construction, and smart city infrastructure.

Argo AI was a self-driving company backed by Ford and Volkswagen. I worked on the perception pipeline — LIDAR and camera processing in real-time. As the most experienced C++ engineer on the team, I led technical and architectural decisions for the codebase. Optimized massively parallel CUDA algorithms by rethinking CPU-GPU memory transfer patterns, and refactored the monolithic codebase to make it testable across teams.

ArtHustle is an online platform for watercolor painting courses. I designed the entire system from scratch — Go backend with GraphQL API, React/TypeScript frontend, microservice-friendly architecture. Beyond coding, a big part of the role was mentoring other developers and setting engineering standards. I introduced and enforced disciplined TDD practices across the team.

Televic GSP builds communication and information systems for public transport vehicles across Europe. I developed multiple low-level UDP-based network protocols in C++ that handle real-time data exchange between trains and central control systems. These protocols drive the passenger information displays you see inside German trains — showing next stops, delays, connections — and the behind-the-scenes communication that keeps the entire national rail network coordinated.

CryEngine powers titles like Crysis and Hunt: Showdown. I built a flexible, configurable, type-safe signal system using C++ template meta-programming — it allowed subscribing to events with arbitrary function signatures, a pattern used widely across the editor. Developed a multi-threaded job scheduler for parallelizing editor operations, which was key to the 18% performance gain. Also built a version control plugin with complex Qt UI for game asset tracking across large teams.