Skip to content
View osick's full-sized avatar
0 followers · 4 following

Achievements

Achievement: Pair Extraordinairex2Achievement: Pull Sharkx3Achievement: QuickdrawAchievement: YOLO

Achievements

Achievement: Pair Extraordinairex2Achievement: Pull Sharkx3Achievement: QuickdrawAchievement: YOLO

Block or report osick

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
osick/README.md

Oliver Sick

Mathematician by training. I build AI systems for large, business-critical enterprise IT — and these repositories are where I try out what has to hold up there later.

In my work I am owning an AI platform as an architect, enabling large projects with dozen of team with agentic environments for large SaFE Value Streams.

What I work on

Legacy modernization with agents How do you move a legacy system to a modern architecture when the business logic was never documented? In 2025 we did exactly that — 50,000 lines, with agents reconstructing the domain rules from the existing code. These repos are the tooling behind it.

  • archaeocode — automated software archaeology: agents excavate business-readable user stories and dependency maps from COBOL, Smalltalk, Fortran and Pascal

Context for AI assistants The model is rarely the problem. The context is.

  • repo-ctx — MCP server giving AI assistants searchable access to any Git repository: symbols, dependency graphs, architecture analysis across 12 languages
  • ctx.skills — agent skills for what LLMs are notoriously bad at: time, structured knowledge, very large context
  • semrag — local-first enterprise RAG with semantic chunking and retrieval
  • knowledge-mcp — MCP server for structured knowledge management

Verify, don't hope Twenty-five years of owning quality in critical systems leave a mark: what you don't measure doesn't work. And an agent you cannot constrain is not a tool, it is a risk.

  • mcp-zerotrust — zero-trust security gateway for MCP: default-deny capability profiles, human-in-the-loop escalation, audit logging. One Go binary between your AI host and its MCP servers
  • urllm — privacy and security audit for any URL. Deterministic extraction first, LLM second, with a --fail-on gate for CI/CD
  • SpecBench — benchmarking spec-driven development frameworks

Mathematics Mathematics is my education and still my love, now working on the combination of AI and math Currently competing in the Mathematics Distillation Challenge run by the SAIR Foundation — see SAIRmcp. So small outcomes were (more to come)

Chess My passion when I don't work on AI and math. A small selection of my work on it

  • ChessMG — chess move generation at engine speed, with a Python API
  • helpmate-tablebase — exhaustive helpmate tablebases for problem composers, evaluating hundred of billion positions in highly compressed databases

Python · Go · C++ · MCP · RAG and GraphRAG · AWS Bedrock · Neo4j · Qdrant Daily driver: Claude Code, GitHub Copilot, Codex, AWS Kiro.

LinkedIn · Hugging Face

Pinned Loading

  1. urllm urllm Public

    privacy and security audit for every URL. Deterministic extraction first, LLM then, with --fail-on-Gate for CI/CD

    Python

  2. repo-ctx repo-ctx Public

    MCP-Server: searchable access on every Git repo for AI assistans - Symbols, dependency graps, architecture analysis for 12 languages

    Python

  3. archaeocode archaeocode Public

    Automated software archaeology — AI agents (LangGraph + MCP) excavate business-readable user stories and dependency maps from COBOL, Smalltalk, Fortran, Pascal & other legacy code.

    CSS

  4. ctx.skills ctx.skills Public

    A growing toolbox of agent skills that make LLMs measurably better at things they are notoriously bad at: time, structured knowledge, very large context

    Python

  5. ChessMG ChessMG Public

    Chess move generation at engine speed with a Python API — the move generator behind the helpmate tablebase project.

    C++

  6. helpmate-tablebase helpmate-tablebase Public

    Exhaustive helpmate tablebases for chess problem composers — complete through 5 pieces. 636 six-piece classes still need computing; contributions wanted.

    C++