Hi! I'm Mikołaj Kocik, a .NET Developer & AI Engineer.
I build applications with a strong emphasis on scalability and architectural integrity, always evaluating multiple approaches to find the best possible answer to an engineering problem.
"From system design to clean implementation."
Warehouse management system.
Handles PZ/WZ/RW/MM operations, generates PDF documents and stores them in Azure Blob Storage. Includes AI-assisted vector search via Ollama.
Eliminates manual stock tracking. An asynchronous document flow offloads the API — a background worker generates PDFs and saves metadata to the database. Deployed via GitHub Actions with OIDC, no stored secrets.
Cloud-native vehicle reservation platform.
Covering the full rental lifecycle: search, booking, Stripe payments, PDF/Excel reporting, and email notifications via Hangfire background workers.
Decouples long-running tasks (reservation deadlines, emails) from the main API thread. Full observability stack with Grafana and Prometheus. Performance verified under load via Azure Load Testing.
Hotel management system.
Includes a full reservation cycle, dynamic availability checks, discount codes, Stripe payments, and guest reviews. Includes a dedicated MCP server for AI agent integration.
Automates guest service via AI agents — an LLM invokes MCP tools (e.g. notify_staff → Slack) without human involvement. Eliminates manual booking errors and handles payments without storing sensitive card data.
Retrieval-augmented documentation assistant.
Indexes large document corpora and serves contextual answers grounded in source material, combining on-device inference with managed Azure AI services.
Reduces time spent searching internal documentation. A hybrid retrieval pipeline (Azure AI Search + ONNX re-ranking) keeps answers grounded and verifiable, while Hugging Face models handle local embeddings.
Intelligent IT helpdesk system.
Combines local LLMs with proactive infrastructure management. Features a real-time chat interface (SignalR), semantic search via pgvector, automated asset discovery through LDAP/Active Directory, and Jira integration for ticket management.
Reduces repetitive support tickets through AI automation. A built-in PII scrubber sanitizes queries before they reach the model, and an ONNX re-ranker via Semantic Kernel keeps retrieval precise. Self-healing logic monitors disk space and restarts services without human intervention.
More projects and descriptions available on my Github profile
GithubDecember 2025 – January 2026
Thinq Sp. z o.o. · Poznań
Computer Science
WSB Merito University in Poznan
