Skip to content
View rjgeng's full-sized avatar
  • San Francisco Bay Area, CA

Organizations

@BloomTech-Labs

Block or report rjgeng

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.

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
rjgeng/README.md

Rongjun Geng

AI/LLM & systems software engineer — agentic systems, RAG, full-stack. San Francisco Bay Area.

I build LLM-powered products end to end — FastAPI + LangChain backends, React frontends — and contribute fixes upstream to open-source multi-agent orchestration systems.

🔀 Merged upstream contributions

Fixes merged across the gastownhall ecosystem — an open-source multi-agent orchestration platform — each reviewed and approved by independent maintainers. 21 merged PRs across 2 repos:

gastownhall/gascity

  • #4406 — fix(cmd/gc): gc init --skip-provider-readiness accepts any builtin provider
  • #4442 — fix(cmd/gc): gc hook --claim never adopts mail message beads as work
  • #4248 — fix(doctor): warn when beads-store silently fell back to BdStore
  • #4219 — fix(runproj): warming snapshot no longer marshals historicalLanes/recentChanges as null
  • #4216 — fix(cmd/gc): lint no longer flags the documented pool-disable form as a named-session conflict
  • …and 15 more — see all 20

gastownhall/gastown

  • #4173 — fix(memories): tolerate non-string values in bd kv list

Themes: concurrency races, operational guardrails, and failure-mode UX in long-running agent systems.

Maintainer on #3469: "a careful, surgical fix — I especially appreciate that you keyed the dedup signature on the set of active conditions."

I also file actionable bug reports that maintainers act on — e.g. #2814 (bundled Dolt 2.0.8 corruption → version pin) and #2846 (compactor quarantine race starving GC) — both closed with a by-name thanks from a core committer.

Stacks

  • Languages: Python, C/C++, Bash, Rust, JavaScript
  • LLM Tools & Frameworks: Hugging Face, LangChain, Gradio, FAISS, Chroma, Ollama, OpenAI, Claude, DeepSeek, Gemini
  • GenAI Techniques: RAG, QLoRA, Prompt Engineering, Agents, Fine-tuning, Multi-modal Integration
  • Backend: FastAPI + LangChain/OpenAI SDK
  • Backend & System: Linux (gdb, perf, valgrind), REST/gRPC/GraphQL APIs, multithreading, system calls
  • Database: PostgreSQL + Chroma (or MongoDB for quick prototyping), MySQL, Redis, SQLite
  • Storage: S3/Supabase for files
  • Frontend: React + TailwindCSS
  • DevOps & Cloud: Docker, GitHub Actions, Jenkins, AWS (S3, Lambda, EC2), Nginx
  • Networking & Security: TCP/IP, WebSockets, MQTT, SELinux, HAProxy
  • Embedded & EDA: Altium Designer, Vivado, Verilog/VHDL, GNURadio, UHD, MATLAB
  • Version Control: Git, GitHub, GitLab, SVN

Pinned projects

See pinned repos below — an AI trading workstation (FinAlly), a private-corpus RAG + web-search assistant (web_rag_search), GenAI/RAG notebooks (gen_ai), and an API-cost postmortem with guardrails. I publish the misses, not just the wins.

Pinned Loading

  1. gen_ai gen_ai Public

    Generative AI notebooks — RAG, prompting, and LLM integration experiments

    Jupyter Notebook

  2. web_rag_search web_rag_search Public

    AI personal assistant suite — private-corpus RAG system + real-time web search agent

    Python