I lead engineering at Touchine, where I'm building OJO—an AI-native design workspace that helps teams turn open-ended product ideas into researched, planned, editable, and production-ready web experiences.

My work sits at the boundary between model capability and production engineering: agent runtimes, workflow contracts, AI gateways, secure execution, distributed systems, databases, reliability, and developer experience.

• Agent systems: planning-first and contract-driven workflows, tool orchestration, memory, evaluation, and human-in-the-loop interaction.
• AI infrastructure: workload-aware model routing, prefix and dynamic caching, observability, cost control, and multi-tenant gVisor sandboxes.
• Distributed systems and databases: vector search, data pipelines, shard rebalancing, diagnostics, performance engineering, and safe migrations.
• Internal AI systems: turning product specifications, engineering architecture, and operational knowledge into reusable skills and workflows.

• OJO: leading the engineering organization across the infinite canvas, agent runtime, platform services, AI gateway, sandbox infrastructure, reliability, and commercialization.
• LUTRA.ICU: my writing and research space.
• Elias: a Cloudflare-native GitHub App for keeping living documentation aligned with source changes.

At Tencent, I worked on vector databases and agent systems, including Agent Memory, Deep Research, multi-vector search, dynamic shard rebalancing, diagnostics, and document ingestion.

Earlier, I worked on database kernels at Huawei GaussDB and contributed to open-source projects including CloudWeGo Netpoll, Apache ShardingSphere, and the CNCF Glossary.

• Portfolio and writing: lutra.icu
• LinkedIn: linkedin.com/in/jacob953
• Email: i@jacob953.com