I am a Computer Science student at Xiamen University, currently focused on backend engineering, AI full-stack development, and practical AI systems.

My interests sit at the intersection of engineering depth and product thinking. I enjoy building backend services, developer tools, and AI-powered workflows that solve real problems instead of stopping at demos.

Right now, I am especially interested in backend development, distributed systems, full-stack AI applications, and turning frontier AI capabilities into usable products.

✦ Email: dengzitao888@163.com
✦ Base: Shenzhen, China

keel-agent · a coding agent CLI that treats sessions as append-only event logs and completion as a verifiable contract (evidence + exit codes for CI). Zero runtime dependencies; OpenAI / Anthropic / Responses protocols.

Merged contributions:

• TencentCloud/Octop (~6.5k ★, self-hosted multi-agent assistant) — 9 of 21 PRs merged, covering OAuth callback security, multi-agent failure semantics, backup import, and connector features
• TencentCloud/octop-harness (Octop's agent runtime) — streamed thinking block projection fix
• TencentCloud/TencentDB-Agent-Memory (~27.7k ★, agent memory hub) — fixes and features, incl. preserving Claude Code AGENTS.md instructions

Under review: L1 memory usage-feedback reinforcement (retrieval practice + forgetting curve reranking), L1 evidence-chain maintenance in TencentDB-Agent-Memory; diagnostics export and parallel-task conflict preflight in Tencent/BrowserSkill; bounded internal sub-agent batches in apache/flink-agents.

I write up the engineering details on my blog: dangzitou.github.io

• Building stronger backend foundations
• Exploring AI full-stack applications and developer workflows
• Following frontier AI progress and testing real product use cases
• Making tools that are practical, local-first, and inspectable
• Contributing to open source through automated agent workflows

Last updated: 2026-10-04