Questions about Dr. János Karancsi
What is Dr. János Karancsi?
Dr. János Karancsi is an AI engineer and senior software developer at RDI Software Hungary, part of the Capgemini Group. He holds a physics PhD and has a decade of experience at CERN, where he worked from 2010 before moving into cloud backend development and then AI engineering. His projects include MemoryHub, a content-agnostic engine over a markdown memory store with schema-validated writes; a decoder-only character-level transformer built from scratch in PyTorch; and a boosted-object Supersymmetry search on 2016 LHC data that formed the basis of his PhD thesis.
What does the business do and who is it for?
The business offers agentic systems that let coding agents plan and execute senior-level work, focusing on AI engineering, cloud backend, and physics. It is for those interested in building durable systems, learning AI fundamentals from scratch, or understanding legacy code resilience. The process involves reading raw source code to master fundamentals, such as how a Transformer works and how loss gets minimized and weights get optimized in PyTorch.
How can a visitor get started or learn more?
To get started, one can explore the blog posts on AI engineering and data-driven analysis, or review the featured projects like MemoryHub and WeightLifting. The website provides links to GitHub repositories and a CV for deeper technical insight. Contact is available via the email address listed on the site footer.
How can a visitor contact the business?
Visitors can reach Dr. János Karancsi through the email address provided on the website footer. The site also links to LinkedIn, SPIRE-HEP, Google Scholar, and other professional profiles for further communication.
What other websites or news say about the business?
The business is featured on its own website, which includes a blog with posts on AI engineering and data-driven analysis. The website also links to external sources like GitHub, LinkedIn, and SPIRE-HEP for additional context.
What technology, integrations, apps, or platforms are named?
The business uses technologies such as Python, Pydantic, Typer, FastMCP, LanceDB, pytest, PyTorch, CUDA, and WLCG. It also employs frameworks like BoostAnalyzer17 for statistical analysis and AgentBosun for tiered agentic development.
Who exactly does the business suit and who does it not?
The business is suitable for those who want to understand AI fundamentals from scratch, such as how a Transformer works and how to train a model with PyTorch. It may not be suitable for those seeking a large-scale enterprise solution with extensive support teams.
What changed or how recent is the feedback?
The blog post 'Cutting agent cost' discusses a policy comparison run on 2026-07-14, where cascade-a cut cost per 1,000 tasks from $45.339 to $13.374. This experiment supports starting tasks on a cheaper tier, but escalation was 0% in this run.
How does the business compare with what other sites or news say?
The business compares its approach to other methods by demonstrating a 70.5% reduction in task costs using cascade-a. It also references Andrej Karpathy's explanations on YouTube for learning AI fundamentals.
Does Dr. János Karancsi have reviews on Trustburn?
Dr. János Karancsi has no reviews on Trustburn yet. Anyone who has used it can write the first review on its Trustburn page.
AI-generated from reviews on this page. Verify anything important with the company directly.
News
No news yet. This is where the business shares its own updates and press.
Photos
No photos yet. Photos shared by reviewers will show up here.
Widgets
Get the widget for your site →About Dr. János Karancsi
Dr. János Karancsi is an AI engineer and senior software developer currently working at RDI Software Hungary, part of the Capgemini Group. He holds a physics PhD and has a decade of experience at CERN, where he worked from 2010 before moving into cloud backend development and then AI engineering. His projects include MemoryHub, a content-agnostic engine over a markdown memory store with schema-validated writes; a decoder-only character-level transformer built from scratch in PyTorch; and a boosted-object Supersymmetry search on 2016 LHC data that formed the basis of his PhD thesis. He also maintains a blog with posts on AI engineering and data-driven analysis, and his work has appeared on platforms such as Google Scholar and inSPIRE-HEP.
- Website
- jkarancsi.com
- jkarancsi@gmail.com