Skip to content
View dvd233's full-sized avatar
🎯
Focusing
🎯
Focusing
  • Beijing, China
  • 13:50 (UTC +08:00)

Block or report dvd233

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
dvd233/README.md
Hi, I'm dvd. AI Full-Stack Engineer.

Looping preview of a two-person team's private hyperspectral reconstruction prototype: camera capture, 30-band reconstruction from 400 to 690 nanometres, and a sampled spectral curve.

Click the preview to open the full muted demo. 10.75 s, 1920x1080, 60 fps, 400-690 nm.

What the demo shows

This is an earlier private hyperspectral reconstruction prototype, co-developed in a two-person team. The public loop and full demo show a desktop workflow for camera acquisition, lightweight deep-learning reconstruction, 30-band spectral display across 400-690 nm, per-point spectral curves, and result saving.

I was responsible for the lightweight deep-learning reconstruction and dataset work, plus the camera acquisition, desktop UI, spectral visualization, and result-saving modules. Code and data remain private because of research and industry-collaboration constraints.

I build reliable AI applications end to end, connecting product interfaces, agent workflows, APIs, and data systems. Computational imaging is my technical edge.

AI applications   frontend + client systems   agent workflows   computational imaging

About

I turn AI capabilities into products people can use: streaming interfaces, local client workflows, orchestration, APIs, and data delivery. My main lane is AI full-stack engineering, with computational imaging and open-source collaboration adding technical depth.

Engineering focus

  • AI application engineering: turning LLM and agent capabilities into product flows across UI, orchestration, APIs, and data.
  • Frontend and client engineering: building streaming interfaces, desktop workflows, device-facing controls, and cross-platform behavior.
  • Agent workflows and data systems: working with RAG, DAG execution, model routing, MCP, SSE, RBAC, state, and storage.
  • Computational imaging: turning reconstruction research into usable software through camera acquisition, spectral visualization, and result handling.

Tech stack

Tech stack across interface, AI applications, backend and data, client delivery, and spectral imaging: HTML5, CSS3, JavaScript, TypeScript, React, React Flow, Zustand, Vite, Node.js, NestJS, Python, PyTorch, LangGraph, RAG, MCP, SSE, WebSocket, PostgreSQL, pgvector, Redis, Rust, Tauri, Git, Windows, PyInstaller, and camera workflows.

  • Interface: HTML5, CSS3, JavaScript, TypeScript, React, React Flow, Zustand, Vite, Webpack, and Rspack.
  • AI applications: LLM agents, RAG, LangChain, LangGraph, MCP tool calling, model routing, SSE, and WebSocket workflows.
  • Services and data: Node.js, NestJS, Python, PyTorch, PostgreSQL, pgvector, Redis, DAG execution, RBAC, snapshots, and structured diff or rollback.
  • Client delivery: Rust, Tauri, Git, Windows tooling, PyInstaller, camera workflows, and desktop result handling.

AI coding workflow

My coding-agent toolkit includes Kimi Code, Codex, Pi, Claude Code, Hermes, OpenCode, and ZCode. I also track DeepSeek Harness as part of the surrounding agent tooling ecosystem. I use these tools inside a human-led loop: understand the codebase, implement the change, test edge cases, document decisions, and review the result.

Official Kimi Code agent mark from Moonshot AI Kimi CodeOfficial OpenAI mark used for Codex CodexOfficial Pi coding agent mark from pi.dev PiOfficial Anthropic mark used for Claude Code Claude Code

Official Nous Research mark for the Hermes agent HermesOfficial OpenCode coding agent mark OpenCodeOfficial ZCode plugin ecosystem mark from Z.ai ZCodeOfficial DeepSeek mark for the DeepSeek Harness project DeepSeek Harness

Open-source contributions

I work on real issues across AI agent systems, desktop UX, web tooling, civic data, and cross-platform behavior.

Contribution map

  • Civic data delivery · open-city-ai/haidian · 27 PRs · 26 merged · 1 open
    Structured submissions, geometry sources, deterministic artifacts, and review semantics.
  • AI agent systems · agentscope-ai/agentscope · 1 open
    Windows subprocess behavior in an AI agent example.
  • Agent memory and Node.js · TencentCloud/TencentDB-Agent-Memory · 3 open
    Proxy preservation, SQLite filtering, and self-healing store state.
  • Code review tooling · alibaba/open-code-review · 1 open
    Go configuration and Jinja template support for review workflows.
  • Desktop UX · desktop/desktop · 1 open
    React and TypeScript pull-request state and suggestion behavior.
  • Web tooling · parcel-bundler/parcel · 1 open
    Correct UTF-8 BOM JSON parsing in the resolver path.
  • Search and coordinates · organicmaps/organicmaps · 1 open
    C++ search behavior for space-separated DMS coordinates.

Merged proof

Complete delivery · bilingual formal submission · merged

Problem: A civic-data submission needed a complete, reviewable package rather than an unsupported concept.

Contribution: I delivered a scoped 41-file bilingual package spanning narrative, offline HTML/PDF, GeoJSON, metrics, sources, a manifest, and deterministic self-checks, while marking provisional geometry and non-endorsement boundaries.

PR: open-city-ai/haidian#2540 · merged for repository intake, not gallery publication or government endorsement.

Site provenance · merged

Problem: Provisional site inputs lacked consistent dates, hashes, licence and use boundaries, while an empty constraints layer could be mistaken for proof that no controls existed.

Contribution: I added a built-ins-only audit for four submitted geometries, recorded six missing official-control layers, and made replacement triggers and permitted uses traceable across the package. The merged change passed 18/18 audit checks.

PR: open-city-ai/haidian#3525

Windows portability · import and worker locking · merged

Problem: An unconditional fcntl import stopped two test modules from being collected on Windows, and worker-lock contention needed consistent cross-platform semantics.

Contribution: I reproduced the failure, introduced the guarded POSIX and Windows lock path, and added the initial regression. Maintainer review then corrected the production file-open path and strengthened its test; the previously uncollectable target suite passed 21/21 tests.

PR: open-city-ai/haidian#3859

Deterministic artifacts · LF across platforms · merged

Problem: Three writer paths emitted CRLF on Windows, creating whole-file diff noise and breaking byte-hash consistency after line-ending normalization.

Contribution: I made the writers emit LF bytes and added direct plus end-to-end regressions. After review exposed a Python 3.9 compatibility issue, I revised the implementation and revalidated both flows.

PR: open-city-ai/haidian#3915

Review semantics · blockers versus future follow-ups · merged

Problem: A flat review-action list converted future conditions into immediate change requests, creating an intake loop that compliant edits could not close.

Contribution: I introduced explicit non-blocking follow-ups with trigger and owner fields, invalidated legacy review caches, and kept current or malformed repairs fail-closed with focused regression coverage.

PR: open-city-ai/haidian#3982; the original blocked head was re-reviewed and merged without weakening existing gates.

Latest merged follow-through: open-city-ai/haidian#4114, add a machine-readable implementation-readiness handoff register and fail-closed audit.

In review now

AI agent systems · 4 open PRs
Frontend, client, and tooling · 4 open PRs
Civic data workflow · 1 open PR
  • open-city-ai/haidian#4106: add a CJK font screenshot regression for the rendered artifact flow; currently in upstream review.

Education

Master's student at BUPT, Electronic Engineering and Information Photonics
Undergraduate background in Network Engineering at BUPT

Contact

dvd@linux.do · z1403594118@gmail.com

Animated GitHub contribution snake generated from dvd233's public contribution calendar, with theme-aware spectral colors.

Animated SVG generated from the public contribution calendar by a scheduled Platane/snk workflow.

Spectral trace footer artwork for dvd's profile, with a contribution-inspired trail on an engineering grid and a 400 to 700 nanometre axis.

Pinned Loading

  1. agentscope-ai/agentscope agentscope-ai/agentscope Public

    Build and run agents you can see, understand and trust.

    Python 30.3k 3.4k

  2. TencentCloud/TencentDB-Agent-Memory TencentCloud/TencentDB-Agent-Memory Public

    TencentDB Agent Memory is a team-level memory hub for AI Agents — turning conversations, docs, and code into four reusable memory assets (Chat Memory, Skill, LLM-Wiki, Code-Graph) that are governed…

    TypeScript 25.5k 2.4k

  3. desktop/desktop desktop/desktop Public

    Focus on what matters instead of fighting with Git.

    TypeScript 21.8k 10.5k

  4. parcel-bundler/parcel parcel-bundler/parcel Public

    The zero configuration build tool for the web. 📦🚀

    JavaScript 44k 2.3k

  5. alibaba/open-code-review alibaba/open-code-review Public

    Fast, efficient, battle-tested at Alibaba's scale. Hybrid architecture code review tool: deterministic pipelines + LLM Agent, precise line-level comments, built-in multi-language ruleset (NPE, thre…

    Go 21.7k 1.6k

  6. open-city-ai/haidian open-city-ai/haidian Public

    HTML 406 1.4k