Mar 2026 – Present · 1 mo · Dubai, United Arab Emirates · Remote

• Backend:
• Implemented NAT-traversal architecture for distributed nodes by shifting connectivity to an outbound persistent tunnel model — built full lifecycle including node registration, keepalive, reconnection handling, and command routing, enabling nodes on home networks, WSL2, or cloud VMs to operate without port forwarding
• Designed a batched workload dispatch system with a time-window aggregation layer, consolidating sequential dispatches into vectorized calls to improve throughput by ensuring multiple tasks arrive within the same aggregation window
• Developed a scalable fleet-wide health check system for distributed nodes, optimizing performance by skipping known-unreachable nodes and running checks in parallel — keeping response time stable as the fleet scales
• Identified and resolved a Go stdlib URL decoding behavior that silently mutated encoded path segments before handler execution; re-implemented routing using the raw path to correctly handle identifiers containing encoded delimiters
• Engineered a real-time crypto analytics pipeline (Rust + PostgreSQL) aggregating multi-exchange data into a unified API, powering interactive dashboards and a Telegram Mini App for live asset, volume, and market insights
• Infrastructure:
• Deployed and self-hosted GitLab for remote teams, secured behind a WireGuard VPN for private, access-controlled code collaboration
• Set up self-hosted Bitwarden for secure, team-wide credential sharing — replacing ad-hoc secret sharing with a centralized vault
• Deployed backend services to production including inference pipeline infrastructure, managing end-to-end rollout and service reliability

Dec 2025 – Mar 2026 · 3 mos · Dubai, United Arab Emirates · Remote

• Designed and built core backend services for a distributed compute orchestration platform, enabling secure interaction with heterogeneous compute nodes and real-time awareness of accelerator availability and load.
• Implemented resilient liveness and state management using heartbeat-driven signals, in-memory caching, and stability logic to ensure consistent interaction with GPU-backed nodes despite network instability or transient failures.
• Developed scalable APIs and persistence layers to ingest execution and resource signals from GPU-powered systems, supporting low-latency routing and orchestration decisions in production environments.