Mateusz OzimkiewiczMateuszAI Automation · Full-Stack

I BUILD AI AUTOMATIONS AND FULL-STACK PRODUCTS

AI Automation & Full-Stack Developer

Practical AI automation, internal tools and API integrations for businesses. Start with a scoped workflow, build a prototype, and keep human review explicit.

Explore solutionsView selected work

WHAT YOU CAN COMMISSION

  • AI-assisted workflows & automations
  • Internal tools & dashboards
  • Document intelligence / RAG

Selected systems

Each project proves a different kind of work you can commission — projects that are concepts or previews are labelled as such.

Yuzuki

Local-first AI desktop with fail-closed private mode

A complete AI product, not a UI mockup: a Tauri desktop application with a live-verified local chat path, seven named routing modes, and Hermes Agent integration. Source link withheld pending lineage review; no binary release.

Tauri · React · TypeScript · Python · Hermes Agent · Ollama

PRIVATE · LIVE-VERIFIED

Yuzuki desktop application showing the session rail, an empty conversation pane, and the context panel.
Yuzuki desktop workspace — synthetic public fixture, empty state.

Klaus — AI Developer Workbench

Safety-first desktop developer workbench

A Tauri / Rust and React / TypeScript desktop workbench with read-only workspace capabilities and an external runtime boundary. Agent run execution remains intentionally blocked.

Tauri 2 · Rust · React · TypeScript · Read-only workspace

IN DEVELOPMENT

MellyCore AIOS

Command-center architecture for shared context and agent coordination

An AI orchestration architecture and shared-context command center, with a graph-centric Portfolio V1 design target shown separately from implemented static-product evidence.

Context Graph · Static product UI · Design target · Safety contracts

ACTIVE DEVELOPMENT · DUAL-SOURCE EVIDENCEDESIGN TARGET · NO LIVE SOURCE

OpsPilot AI

Human-in-the-loop AI operations automation system

An evidence-led operations console for deterministic lead qualification, recommended actions, human approval, safely simulated execution, and audit history.

React / TypeScript / FastAPI / PostgreSQL / Human approval

FRONTEND + POSTGRESQL / ACCEPTEDDEMO DATA / SIMULATED / NO LIVE OUTBOUND

RAG Assistant

RAG-powered document search with source-grounded answers

Document intelligence for knowledge bases, procedures and support documentation: questions answered from a company's own documents, each answer shown with its sources.

Retrieval · Embeddings · Indexing · Citations

IN DEVELOPMENT · CONCEPT INTERFACE

Other work

MellyTrade

Data-dense analytics and monitoring dashboard for signals, risk and system health

An implemented full-stack market analytics application with a FastAPI backend and React/TypeScript frontend. Read-only and paper-safe, with no broker or live order execution.

FastAPI · React · TypeScript · PostgreSQL / Supabase · Docker · Tauri / PWA · GitHub Actions

FULL-STACK · IMPLEMENTEDREAD-ONLY · PAPER-SAFE

AI Agent Workspace

Documented multi-agent development workflow

A documented workflow connecting ChatGPT, Claude Code, Codex, Obsidian and GitHub through reusable context, task contracts and review gates.

Claude Code · Codex · Obsidian · GitHub · Documentation

What I build

The categories of work I take on, each tied to evidence on this site.

AI automation & agents
Agent workflows · Task routing · Human review · Context handoff
Evidence: Yuzuki's routing and fail-closed provider check
Document AI / RAG
Ingestion · Chunking · Embeddings · Retrieval · Source-cited answers
Evidence: RAG Assistant (concept interface)
Full-stack products & APIs
React · TypeScript · Python · FastAPI · Tauri
Evidence: The Yuzuki desktop product
Dashboards & internal tools
Data-dense UI · Monitoring · Risk and status views
Evidence: MellyTrade (implemented full-stack, read-only / paper-safe)
Testing & validation
pytest · Playwright · GitHub Actions · Documentation
Evidence: Yuzuki's public validation suites (raw counts)

Knowledge Atlas

Five projects and the engineering capabilities they share. Every line is a declared relationship — what a project is built with, or what its design requires. A shared capability is not a shared runtime: no line joins two projects directly.

  • BUILT_WITHshown in implemented work
  • REQUIRESpart of a concept or architecture — not an implemented claim
  • Static map · authored data · not a live graph
Knowledge Atlas — five projects and the six engineering capabilities they shareFive project nodes sit above and below a horizontal band of six capability nodes. Each line is a declared relationship from a project to a capability: solid for built with, dashed for requires. No line joins two projects directly, because the projects do not share a runtime.YuzukiACCEPTED PRODUCTMellyCore AIOSIMPLEMENTED · NOT LIVEOpsPilot AIIMPLEMENTED · NOT LIVERAG AssistantCONCEPT INTERFACEMellyTradeIMPLEMENTED · NOT LIVERAGKnowledge & contextAgentsSafety & governanceAutomationData & analytics
  • 3 relationships: built with Agents; built with Knowledge & context; built with Safety & governance.
  • 3 relationships: requires Agents; built with Knowledge & context; built with Safety & governance.
  • 3 relationships: built with Automation; built with Safety & governance; built with Data & analytics.
  • 2 relationships: requires RAG; requires Knowledge & context.
  • 2 relationships: built with Data & analytics; built with Safety & governance.
  • 1 relationship: RAG Assistant (requires).
  • 3 relationships: Yuzuki (built with); MellyCore AIOS (built with); RAG Assistant (requires).
  • 2 relationships: Yuzuki (built with); MellyCore AIOS (requires).
  • 4 relationships: Yuzuki (built with); MellyCore AIOS (built with); OpsPilot AI (built with); MellyTrade (built with).
  • 1 relationship: OpsPilot AI (built with).
  • 2 relationships: OpsPilot AI (built with); MellyTrade (built with).

Select a project to see its maturity, the evidence behind each relationship, and a link to its case study. Select a capability to see which projects share it.

  1. Yuzuki

    ACCEPTED PRODUCT

    Local-first AI desktop with fail-closed private mode

    PRIVATE path live-verified; six cloud routing modes implemented, not live-verified.

    • BUILT_WITHAgentsHermes Agent integration.
    • BUILT_WITHKnowledge & contextSession persistence and local search; the knowledge surface is captured as a prototype on fixture data.
    • BUILT_WITHSafety & governancePRIVATE fail-closed provider check, live-verified.
    Open the Yuzuki case study

  2. MellyCore AIOS

    IMPLEMENTED · NOT LIVE

    Command-center architecture for shared context and agent coordination

    Static cockpit implemented and deployed as a static preview; no live runtime, providers or running agents.

    • REQUIRESAgentsAgent-coordination architecture; routing contract accepted at specification level only. No agents run.
    • BUILT_WITHKnowledge & contextDeterministic, repository-backed knowledge and operations graph on fixture data. Not live.
    • BUILT_WITHSafety & governanceExecution-locked static cockpit, supervised by design.
    Open the MellyCore AIOS case study

  3. OpsPilot AI

    IMPLEMENTED · NOT LIVE

    Human-in-the-loop AI operations automation system

    Frontend built and accepted on fixture data; approval and execution are simulated, with no live outbound.

    • BUILT_WITHAutomationQualification → recommended action → approval → safely simulated execution.
    • BUILT_WITHSafety & governanceExplicit human approval before anything executes.
    • BUILT_WITHData & analyticsAudit history; PostgreSQL 16.15 technical acceptance on a disposable instance.
    Open the OpsPilot AI case study

  4. RAG Assistant

    CONCEPT INTERFACE

    RAG-powered document search with source-grounded answers

    Interface concept only; no retrieval backend, corpus or deployment is claimed.

    • REQUIRESRAGArticulated retrieval architecture: documents → chunking → index → retrieval → grounded answer.
    • REQUIRESKnowledge & contextDesign intent: answers drawn from a company's own documents, each shown with its sources.
    Open the RAG Assistant case study

  5. MellyTrade

    IMPLEMENTED · NOT LIVE

    Data-dense analytics and monitoring dashboard for signals, risk and system health

    Implemented full-stack market analytics application; read-only and paper-safe, with no broker or live order execution.

    • BUILT_WITHData & analyticsFastAPI and SQLAlchemy backend, React/TypeScript frontend, and a read-only PostgreSQL/Supabase data layer.
    • BUILT_WITHSafety & governanceRead-only and paper-safe by design; live order execution is blocked and no broker is integrated.
    Open the MellyTrade case study

Shared capabilities

RAG
RAG Assistant
Knowledge & context
Yuzuki · MellyCore AIOS · RAG Assistant
Agents
Yuzuki · MellyCore AIOS
Safety & governance
Yuzuki · MellyCore AIOS · OpsPilot AI · MellyTrade
Automation
OpsPilot AI
Data & analytics
OpsPilot AI · MellyTrade

Find Klaus — AI Developer Workbench and AI Agent Workspace in theSystems index.

Watch a request meet the router.

In private mode it hits a hard stop instead of a cloud fallback.

PRIVATE · LIVE-VERIFIED6 MODES · IMPLEMENTED · NOT LIVE-VERIFIED

Enter the Yuzuki case study

ACTIVE DEVELOPMENT · DUAL-SOURCE EVIDENCE

MellyCore

A systems architecture for agent orchestration: an operating loop where a human approves before anything is implemented. It is not a shipped public product.

Read the MellyCore case study →

MellyCore operating loopA closed loop of seven steps: observe, analyse, recommend, approve, implement, validate, record, then back to observe. Approve is the one emphasised step, labelled human in the loop.observeanalyserecommendapproveHUMAN-IN-THE-LOOPimplementvalidaterecord
  1. observe
  2. analyse
  3. recommend
  4. approveHUMAN-IN-THE-LOOP
  5. implement
  6. validate
  7. record
  8. ↺ back to observe

Implemented

  • Cockpit V3.1 static product
  • Production-deployed static preview
  • Execution locked — no live runtime

Design target

  • Claude Design MellyCore AIOS v2
  • Portfolio V1 graph-centric command center
  • Static preview — no connected live source

Specification accepted

  • Skill/plugin routing contract v1.2
  • Accepted at specification level
  • Not an operational runtime

Simulated: model-comparison copy is deterministic local text, not live model responses.

FRONTEND + POSTGRESQL · ACCEPTEDDEMO DATA · SIMULATED · NO LIVE OUTBOUND

OpsPilot AI

A human-in-the-loop operations console that turns lead evidence into traceable qualification, recommended actions, approval decisions, and safely simulated execution.

  1. Inbound lead
  2. Analysisevidence-led
  3. QualificationHOT / WARM / LOW
  4. Recommended action
  5. Human approval — key stage
  6. Simulated executionno live outbound
  7. Audit evidence
OpsPilot operations overview with lead qualification counts, action queue, demo-data disclosure, and no-live-outbound boundary.
Accepted OpsPilot overview — deterministic demo data; no live outbound.

Read the OpsPilot case study →

How I build

  1. Discover
  2. Architecture
  3. Safety contract — key stage
  4. Build
  5. Validate — key stage
  6. Real-device acceptance
  7. Ship

Yuzuki's fail-closed provider check is the safety contract; its 23/210/63/7 suites are the validation.

Systems index

A directory view of the same work — not the navigation, just the map.

~/melly

About

AI Automation & Full-Stack Developer working across React, TypeScript and FastAPI, building AI automations and full-stack products for real business workflows — with an AI-assisted workflow built around engineering discipline: dry-run defaults, read-only modes, and honest limits.

More about me →

Have a workflow that still depends on repetitive manual work?

I build AI automation, internal tools and full-stack AI applications for businesses. Send a short description of the process — what happens today and what should happen instead — and I will reply by email.

Discuss your workflowStart with an automation auditSolutions & draft pricing →

Colophon

Mateusz Ozimkiewicz · Available now · Poland / Remote · B2B · Freelance · Contract ·

Every claim this site makes, and how well each is proven:

  1. Local chat path, end to endLIVE-VERIFIED
  2. PRIVATE fail-closed provider checkLIVE-VERIFIED
  3. Six cloud routing modesIMPL · NOT LIVE-VERIFIED
  4. ADVANCED read-only projectionIMPL · NOT LIVE-VERIFIED
  5. Knowledge ConstellationPROTOTYPE / FIXTURE DATA
  6. Lifecycle / Job Object process treeSHIPPED
  7. Validation suitesVALIDATED
  8. Public distributionSOURCE ONLY
  9. MellyCore implemented layersACTIVE DEVELOPMENT
  10. MellyCore Portfolio V1 design targetPROTOTYPE / FIXTURE DATA
  11. MellyCore ObservatoryPLANNED
  12. Klaus — AI Developer WorkbenchIN DEVELOPMENT
  13. RAG Assistant interfaceCONCEPT · NOT YET BUILT
  14. MellyTrade full-stack analytics applicationIMPL · NOT LIVE-VERIFIED
  15. OpsPilot AIVALIDATED