Advanced RAG
High-accuracy knowledge-base retrieval: hybrid dense + sparse search, LLM pruning, grounded answers with factual checks, vector databases and embedding models.
IT & AI Architect · Ph.D. in Artificial Intelligence
I design AI systems that businesses can trust.
Ing. Michal Bureš, Ph.D. — an architect who understands both technology and business. Hands-on with LLMs, retrieval-augmented generation and agentic AI, on top of 20 years of enterprise and solution architecture for major banks and telecom operators.
- Ph.D.AI & Biocybernetics, Czech Technical University
- 7 yrshands-on AI: machine learning, LLMs, RAG, AI agents, neural networks
- 20 yrsenterprise & solution architecture
- 14 yrsbanking, insurance & finance
AI expertise
From research foundations to production AI
My post-graduate degree is in artificial intelligence and machine learning. Today I architect conversational AI end to end — retrieval, orchestration, evaluation and the security that makes it safe in a regulated enterprise.
Agentic AI & tool calling
Agent orchestration and LLM-based scripting of tool calls, including MCP — models that act on real APIs, safely sandboxed and approved.
Evaluation & testing
Automated testing and evaluation of conversations: golden queries for retrieval, metrics over every pipeline stage, regression suites you can run on every change, dynamic test agents with specific personas.
Document intelligence
AI-based document recognition and processing in Microsoft Azure — Azure AI Foundry and Document Intelligence in production workflows.
Secure AI for regulated industries
Banking-grade security applied to AI: OAuth 2 / OpenID Connect, strong customer authentication, role-based access to knowledge, EU data sovereignty.
AI solution architecture
From use case to production: architecture, estimates, vendor and model choices, and technical presentations to executive management.
The real project
Pontra — an enterprise AI chatbot platform
I designed and built Pontra: an industry-agnostic platform for grounded, multilingual AI assistants, ready for EU regulation. The assistant on this page runs on it — ask it how it works.
- Four-wave pipeline with speculative parallelism — context, execution, output, follow-up.
- Hybrid RAG over an LLM-enriched knowledge base, with pruning and a factual check.
- Sandboxed code — the LLM writes JavaScript that calls tools; isolated-vm runs it.
- Live skills — weather, calculations, calendar and charts; tenant APIs via tool services.
- Realtime voice bot over WebRTC, sharing knowledge and skills with the text chat.
- Built to scale — stateless replicas, Redis Streams, load shedding, an eval framework.
1
Contextlanguage · safety · embedding · skills
2
Executionhybrid RAG · orchestrator · sandbox
3
Outputgrounded synthesis · factual check
4
Follow-upsuggestions · charts
Experience
Featured projects
-
AI Architect · Erste Digital (Erste Group)
Prague & Vienna
Architect of an industry-leading chat-bot solution: advanced RAG, API integration using MCP, data visualisation, agent orchestration and automated evaluation of conversations.
-
Solution Architect · Česká spořitelna (Erste Group)
Prague · Housing & mortgages
IT architecture of the mortgages domain; AI-based document recognition in Azure and an automated pricing-exception approval workflow.
-
Solution Architect · HelloBank! (BNP Paribas)
Prague · Payments & digital banking
Digital loan onboarding, KYC and risk scoring, 3DS v2 and NFC payments, PSD2 open-banking API and strong customer authentication.
-
Enterprise integration lead
Prague · Banking, insurance & integration
Led an integration competence centre: pre-sales, ESB platforms, governance, and a PSD2 open-banking API for a major Czech bank.
-
Freelance architect and integration consultant
Banks, an airline, a railway and telecom operators across Europe, the Middle East and Africa
Enterprise and solution architecture for large organisations — the foundation for designing AI that fits real enterprise systems.
Full project list on LinkedIn.
International experience
Delivered on site in nine countries, on three continents
Working inside client teams abroad — from European banks to launching two green-field mobile operators in Africa — taught me to align people across cultures, time zones and very different organisations. That matters as much for AI adoption as the technology does.
- Czech RepublicBanking · AI
- AustriaBanking · AI
- GermanyAir cargo · Rail
- NetherlandsBanking
- United KingdomBanking
- PortugalTelecom
- United Arab EmiratesTelecom
- UgandaTelecom · green-field launch
- Republic of the CongoTelecom · green-field launch
Education
Grounded in research
- Ph.D. in Artificial Intelligence and BiocyberneticsCzech Technical University in Prague · 2003 – 2007
- Ing. (MSc) in Technical CyberneticsCTU, Faculty of Electrical Engineering · 1996 – 2003
Industries
- Banking, insurance & finance14 years · loans, mortgages, payments, cards, digital channels, PSD2
- Telecommunications6 years
Technologies
Tools I build with
- OpenAI GPT
- Embedding models
- RAG
- Vector databases
- Qdrant
- MCP
- Agentic AI
- LLM evaluation
- Azure AI Foundry
- Document Intelligence
- Databricks
- Redis
- Node.js
- TypeScript
- Python
- Kubernetes
- Terraform
- OAuth 2 / OIDC
English — proficient · German — basics
Contact
Let's talk about your AI
Prefer a quick answer? The assistant knows my background and how Pontra works. Want to try its voice mode? Write me for a personal access link.
- E-mailmichal@michalbures.eu
- Contact cardDownload vCard
- LinkedInlinkedin.com/in/michalbures
- Based inPrague, Czech Republic (EU)