Zainul Masadi

I trained and evaluated Grok-Code-Fast-1 at X.AI using RLHF, improving inference throughput from 180 to 200 tokens per second. I designed reward-signal criteria for code reasoning and applied distributed-systems experience to inference-serving performance.

At ByteDance, I led architecture for the Tokopedia x TikTok Shop fulfillment integration, supporting 2x GMV growth. I designed cross-region systems targeting 99.99% availability and 25K+ RPS, while mentoring up to 12 engineering teams on scalable design and observability.

At Tokopedia and GoTo Logistics, I scaled logistics platforms from 500K to 1.5M+ daily orders, delivered $8M+ in annual infrastructure savings, and built warehouse automation handling 150K+ daily orders. I also developed a packaging-recommendation engine that reduced shipping costs by $125,000 per month.

I'm now building Guardi, a vendor-agnostic AI SDLC orchestration platform combining multi-LLM workflows, RAG, persistent memory, sandboxed execution, and automated security governance. I enjoy designing reliable AI-native systems that help engineering teams ship faster and operate at scale.