Description

Founder of MadArtSoft Ltd. and senior software engineer with 25+ years building production systems — from 8051 embedded firmware to multi-tenant AI SaaS on Azure.

What I do for clients:

• AI / LLM engineering — RAG pipelines, agentic systems, citation verification, privacy-by-design anonymisation. I built LexTool, a legal-opinion SaaS for Polish law firms using hybrid OpenSearch + ChromaDB retrieval and a LangGraph orchestration layer.

• Systems integration — identity and access control bridges (SALTO KS ↔ Microsoft Entra ID, Workday ↔ Symmetry), IoT and CAN-bus telemetry pipelines, and access-control platforms unifying Gallagher, Symmetry and Workday under one API.

• Full-stack delivery — C# / .NET 8 backends, React 18 / TypeScript frontends, Azure infrastructure with Bicep and GitHub Actions CI/CD.

I work best when the problem crosses layers — architecture, backend, AI, embedded, frontend, ops — and the bar for correctness is high. Clients hire me to own outcomes, not just write code to spec.

Recent and current sectors: legal-tech, lithium-battery manufacturing, access control, pharmaceutical software.

Remote across UK and EU. Fluent in English and Polish.

Industry field of expertise

Languages

English
Fluent
Polish
Native or bilingual

Workplace preferences

Remote only

Primarily works remotely

MadArtSoft Ltd
Founder / Lead Developer
October 2025 - Today (10 months)
Project:
AI-native SaaS for Polish law firms — RAG-based legal-opinion drafting
• • Designed and built end-to-end — a multi-tenant SaaS that drafts legal opinions and procedural documents grounded in real Polish statutes and case law.
• • AI / RAG service in Python (FastAPI, gRPC): hybrid retrieval over three OpenSearch indices (private docs, statutes, judgments), single _msearch round-trip, merged with weighted Reciprocal Rank Fusion. ChromaDB for article-level vectors that grows automatically as ELI articles are fetched.
• • Citation verification: LLM-backed article normaliser ("764¹" / "764/1" / "art. 764
◦ § 1" → canonical key), act resolver against ELI / ISAP, and a verifier that flags hallucinated or paraphrased citations before the draft reaches the lawyer.
• • Privacy-by-design anonymisation: spaCy Polish NER (pl_core_news_lg) + regex for PESEL / NIP / REGON / IBAN / phone / email, reversible mapping, fail-closed before any outbound LLM call.
• • LangGraph state machine orchestrating multi-turn opinion sessions — missing-articles fetch/retry, failed verification revise, user-feedback re-entry.
• • C# / .NET 8 backend (clean architecture, EF Core, SQL Server) — multi-tenant middleware on JWT claims, EF global query filters, per-org Azure resource group provisioning for premium tenants (own SQL, own Key Vault, own Container Apps).
• • Two React 18 + TypeScript front-ends (lawyer workspace + admin console) — Vite, Tailwind, Zustand, React Query, Auth0.
• • Infra: Azure Container Apps, Azure SQL, Key Vault, Storage, ACR. Bicep templates (subscription-level + group-level) and GitHub Actions for CI/CD across all tenants.
Tech: C# .NET 8, Python 3.11, FastAPI, gRPC, OpenSearch, ChromaDB, sentence-transformers (mmlw-roberta), llama.cpp, OpenAI API, LangGraph, spaCy, EF Core, SQL Server, Auth0, React 18, TypeScript, Vite, Tailwind, Azure
MadArtSoft
Lead Developer
September 2025 - Today (11 months)
Designed and developed BMS Passport, a software product targeting EU Battery Regulation (EU 2023/1542) compliance ahead of the 2027 deadline. Targeted at industrial SMEs, the solution provides a Docker-based, deployable platform for managing battery passport data and reporting.
Key responsibilities:
Defined product strategy, technical architecture and go-to-market plan for the Polish industrial SME market
Developed core backend services and Docker-based deployment packaging
Planned integration roadmap including a Modbus-TCP connector for Victron energy systems (Phase 2)
Produced sales templates, technical documentation and business plan
Tech stack: C#, .NET, Python, Docker, Modbus-TCP, REST APIs
Python C# CAN-BUS IoT
MadArtSoft
Lead Developer
September 2025 - Today (11 months)
RAG-based Legal/Document Processing Platform
Role: Founder / Lead Architect & Developer Duration: [add dates] – ongoing Industry: Legal Tech / Enterprise AI
Designed, built and deployed LexTool — a Retrieval-Augmented Generation (RAG) document processing platform for GDPR-compliant, on-premise and cloud deployments. The system uses a hybrid architecture: a C# / .NET Web API as the main application layer and Python gRPC microservices for AI/RAG workloads.
Key responsibilities:
Architected and developed C# .NET 8 Web API and Python gRPC microservices
Migrated the RAG backend from a Supabase-based stack to a standalone Python microservice with Azure SQL
Deployed and operated the solution on Azure Container Apps, including custom domain (app.lextool.pl), revision-based traffic splitting and a maintenance-mode mechanism via nginx
Built a dedicated MCP server (LexTool.MCP) integrating Claude Code CLI with internal SQL and Container Apps tooling, plus a custom deployment skill
Implemented GDPR-compliant document ingestion, embedding and retrieval pipelines
Tech stack: C#, .NET 8, ASP.NET Web API, Python, gRPC, Azure Container Apps, Azure SQL, Docker, nginx, MCP
Python SQL C#