About

I build software, and I like keeping it simple and secure. Right now I build agents that review research papers at ETH Zürich. Before that, I built multilingual LLM pipelines at Infineon.

I studied computer science at ELTE in Budapest, where I also taught and co-founded the Data Science Club. Here's how I got here →

Full CV ↗Email me ↗

Current work

  • Research Assistant @ ETH Zürich Dec 2025 → present

    Building agents that review research papers, then testing where their feedback holds up and where it falls apart at the Agentic Systems Lab.

Previously

  • Responsible AI Research Intern @ RC2S2 Sep 2025 → Aug 2026

    Built a framework that asked five models the same sensitive questions as 15,552 synthetic personas. The answers changed depending on who the models thought was asking.

  • Software / AI Engineering Intern @ Infineon Technologies Aug 2025 → Jul 2026

    Cegléd · Built a production multilingual LLM extraction pipeline in Python that turns support conversations into 10+ structured fields across 4+ languages, feeding internal ticket routing. Packaged it as a reusable module with Docker, GitHub Actions, logging, and run-level metrics.

  • Machine Learning Research Intern @ ELTE Jul 2025 → Dec 2025

    AI Department · Operationalized ACOM, an ant-colony mapping method for 2D document embedding. It beat PCA, t-SNE and UMAP by +11.6% neighborhood preservation and +3.8% trustworthiness on 20 Newsgroups. Accepted to ACM ICPS.

  • Research Intern @ ELTE Aug 2025 → Nov 2025

    Department of Numerical Analysis · Built a Python pipeline that rejects bad finger-worn sensor data before comparing its heart-rate readings with ECG.

Publications

  • Biologically Inspired Document Embedding in Discrete Spaces ACM ICPS · 2026

    AI research · ELTE AI Department, Prof. László Gulyás · oral presentation at IntRob Fair 2026. Continuous embeddings are useful but hard to inspect. We placed documents on a discrete semantic grid, then checked what structure survived compared with PCA, t-SNE, and UMAP.

  • Boundary-Work by Artificial Intelligence: How Large-Language Models Govern Sensitive Public Discourse Across User Personas STS NL 2026 · ESA 2026

    Responsible AI · RC2S2, ELTE · with Ildiko Barna, Renáta Németh, Márton Gosztonyi. Presented at STS NL 2026, University of Twente; accepted talk at the 17th ESA Conference, Warsaw, manuscript in preparation. We tested the same sensitive questions across 15,552 synthetic personas. The models' answers changed depending on who they thought was asking.

Education

  • ELTE · BSc Computer Science graduated July 2026

    Budapest · Moved from Pakistan to Budapest on a Stipendium Hungaricum scholarship, selected from roughly 40,000 applicants.

Teaching

  • Teaching Assistant @ ELTE Feb 2026 → Jun 2026

    Taught operating systems and data structures to 60+ students. Most weeks meant labs, code reviews, consultations, and a lot of debugging.

Community

  • Co-founder · ELTE Data Science Club Sep 2025 → present

    Kumar and I started with an empty room and no budget. The club grew to 200+ members and five project teams.

  • Student Mentor @ HÖOK

    Helped new Stipendium Hungaricum students handle the first confusing weeks of university life in Hungary.

  • Senior Student @ ELTE Student Support Centre

    Ran mentor classes with a faculty mentor, giving new international students somewhere to bring the questions that never fit an official handbook.

  • GDG Agentic AI Hackathon Budapest 2026

    Co-organized a 150-person agentic AI hackathon. Morgan Stanley, Wise, Loffice, and ElevenLabs joined as partners.