Bencium® Agentic AI
Bence Csernak
| **Applied Agentic AI Engineer and Consultant @ Cognitum |
AI adoption advisor for leaders. I take AI from pilot to daily use your team trusts |
Ex-HSBC, EY |
Co-founder and Board Member, Agentics Foundation** |
AI Architecture, Engineering, Design and Build
I combine practical AI engineering, product design and strategic leadership. I design, deploy, evaluate and improve LLM-based applications: agentic systems, multi-agent orchestration, advanced RAG and conversational AI. My work turns AI from experiments into production systems that handle real user behaviour and operate safely in regulated environments. I help teams understand what AI can do, where it fails, and how to build useful systems with the right governance, evaluation and secure data handling. Applied AI research and product work since 2017.
🧭 AI Adoption for Leaders
- Moving teams from pilots and demos to AI they use every day and trust
- Clear explanation of AI capabilities, limits and risks for executive, technical and non-technical audiences
- Governance, evaluation and human-in-the-loop review that make adoption safe in regulated industries
- Hands-on adoption workshops, enterprise training and upskilling with Claude Code, Codex CLI, Gemini CLI and custom orchestration platforms
🛠️ Technical Expertise
- Agentic AI & LLM Engineering: agentic systems, context engineering, multi-agent orchestration, AI-assisted development workflows
- RAG & Data: embeddings, vector search (Pinecone, Supabase/pgvector, Weaviate, Chroma, Neon), multimodal RAG, retrieval quality testing, source attribution, hallucination reduction, conversation-log analytics
- Evaluation & Governance: evaluation pipelines for accuracy, safety, compliance and business impact; model risk management; secure data handling
- Agent Tooling: MCP servers, Agent Skills (SKILL.md), Claude Code plugins and hooks, Claude Code, Codex CLI, Gemini CLI, RuFlo
- Models: OpenAI, Anthropic, Gemini, Mistral and open-source models
- Development: TypeScript, Node.js, Python, Next.js, React, Supabase, PostgreSQL, Tailwind CSS, shadcn/ui; deployment on AWS, GCP, Azure, Vercel and Fly.io
- Design & UX: product strategy, AI product design, data visualisation, design systems
🧩 Open Source
- bencium-marketplace (450+ ⭐): Claude Code plugin marketplace with 17 skills and tools for design, marketing, architecture and productivity, including the controlled, innovative and impact UX designer skills, design audit, typography, human-architect-mindset and vanity-engineering-review. SKILL.md format, works across 40+ agentic tools via
npx skills add bencium/bencium-marketplace.
- bencium-loop: lightweight build loop for agentic coding. Plan/spec gate, acceptance checklist with an independent verifier, hook-blocked deploys until verification passes, and a project memory that survives context resets. Runs in Claude Code, Codex and opencode.
- Agent Therapy: independent counselling for AI agents that are stuck in loops, facing conflicting instructions or pressured by users. It names the failure state, gives a short recovery plan and says when to hand over to a human. Ships as an MCP server (listed in the MCP Registry), a Claude Code plugin with a watchdog, an Agent Skill for 40+ tools, and npm and PyPI clients.
- agi-detector: experimental detector for early signs of artificial general intelligence emerging.
- More at github.com/bencium.
🚀 Current Work
- Bencium: Founder. AI adoption, agentic engineering and design consultancy, from proof of concept to production
- Cognitum: Applied Agentic AI Engineer and Consultant
- MedHubAI: Co-Founder and CTO. Production conversational agents for clinic operations that cut patient no-shows by about 35%, with evaluation for clinical safety, accuracy, hallucination risk and compliance
- brandguideAI: Co-Founder and Agentic AI Engineer. Conversational brand-building platform with multimodal RAG over 1,000+ PDF pages and 90%+ relevant source attribution
- Agentics Foundation: Board Member, London Chapter Co-Founder, Research Committee Chair, Educator
- AnnaAI: Peer-reviewed non-profit research on an empathic AI assistant for cancer patients
Recent: enterprise AI automation trainer (Hunter Douglas Group).
🏦 Background
15+ years of product and UX leadership across financial services, government and healthcare: HSBC, EY, NatWest Markets, BNY Mellon (via Digiterre), Publicis Sapient and Nortal.
Trusted by teams at EY, HSBC, NatWest Markets, BNY Mellon, Coca-Cola, Publicis Sapient, Nortal, Genesis Energy, Digiterre and SEEN Group.
🎤 Speaking & Teaching
- O’Reilly Media: live online course author and presenter, Building Agent Skills for Claude Code
- AI Summit Budapest: presenter 2026; speaker September 2025, Agentic AI in Healthcare Communication
- AE Global Summit, London (Oct 2025): selected community ambassador
- Service Design Day Budapest (Oct 2025): Designing the full patient journey with AI agents
- PromptUX Berlin (Apr 2025): featured speaker, Security and Compliance in the EU
- Agentics Foundation: workshops, London meetups and hackathons
- Training: Banyan Software agentic AI and context engineering course (6-week remote cohorts of 20–60 enterprise developers, 2026); TechNativeDigital in-person, government-funded AI transformation course for developers; HelloYellow AI Bootcamp; Prjctr Institute design leader AI workshop; SEEN Group staff training
- Research: AnnaAI peer-reviewed paper, indexed in PubMed and Google Scholar
- Earlier: first AI talk “Banking and UX: A Design Revolution” on emotional UI at The Trampery, London (May 2017); Lloyds Bank Innovation Lab and UX Crunch (2017); Techstars mentor (2015)
Writing: weekly Substack on practical AI for builders and leaders, the Bencium Podcast, and LinkedIn. Launched AIdesign.Tools, one of the first AI design knowledge bases (2021, archived).
🤝 Let’s Connect
💡 Open to advisory, consulting and collaboration on AI adoption and agentic systems.