Data Scientist / ML Engineer · Berlin

Roman Kochnev

Hi! I'm a data scientist with an applied math background and industry experience. I like problems where LLMs do more than chat: designing neural networks, reasoning over knowledge graphs, and automating the tedious parts of ML.

I recently finished my M.Sc. in Computer Science at the University of Würzburg and defended my thesis, Enabling Symbolic Reasoning with Advanced RAG Systems. I wrote it at the Bosch Center for Artificial Intelligence (BCAI) in Renningen, Germany, under the supervision of Dr. Daria Stepanova and Dr. Cuong Xuan Chu. I also work as a research assistant at the Computer Vision Lab in Prof. Dr. Radu Timofte's group, mostly on AutoML and LLM-based neural architecture search.

Publications

APOLO: Automatic Prompt Optimization for Ontology Learning

R. Kochnev^*, H. T. Mai^*, C. X. Chu, L. Lange, H. Paulheim, D. Stepanova

ISWC 2026 · co-first author

^* Equal contribution

Paper
NNGPT: Rethinking AutoML with Large Language Models

R. Kochnev, W. Khalid, T. Uzun, X. Zhang, Y. Dhameliya, F. Qin, C. Vysyaraju, R. Duvvuri, A. Goyal, D. Ignatov, R. Timofte

CVPRW 2026 · first author

Paper Code
LEMUR 2: Unlocking Neural Network Diversity for AI

T. Uzun, W. Khalid, S. Din, S. Mulukuledu, A. Singh, C. Vysyaraju, R. Duvvuri, A. Goyal, Y. R. Lukhi, M. A. Hussain, K. Jesani, U. Shrestha, Y. Mittal, R. Kochnev, P. Kadam, M. Ikram, H. Moradiya, A. Arslanian, D. Ignatov, R. Timofte

CVPRW 2026

Paper Code
LEMUR Neural Network Dataset: Towards Seamless AutoML

A. Goodarzi, R. Kochnev, W. Khalid, H. Goudarzi, F. Qin, T. Uzun, Y. Dhameliya, Y. Kathiriya, Z. Bentyn, D. Ignatov, R. Timofte

arXiv preprint, 2025 · second author

Paper Code
Optuna vs Code Llama: Are LLMs a New Paradigm for Hyperparameter Tuning?

R. Kochnev, A. Goodarzi, Z. Bentyn, D. Ignatov, R. Timofte

ICCVW 2025 · first author

Paper
Intelligent Method for Real Estate Price Prediction in Moscow Using a Machine Learning Model

R. Kochnev, R. Senchenko

75th Days of Science of NUST MISIS Students, 2020 · ISBN 978‑5‑907227‑11‑8

Paper

Experience

2026

Data Science Researcher (Master's Thesis)

Bosch Center for Artificial Intelligence (BCAI), Renningen, Germany

LLMs for ontology learning: developed APOLO, a prompt-optimization framework accepted at ISWC 2026; built GRAIL, a reasoning-focused QA benchmark for RAG and GraphRAG; designed OWL2Doc, a multi-agent pipeline for text-ontology data.
2025–present

Research Assistant (HiWi)

Universität Würzburg, Würzburg, Germany

AutoML with LLMs in the Computer Vision Lab: first author of NNGPT (CVPRW 2026); co-developer of the LEMUR dataset (LEMUR 2 paper, CVPRW 2026); first author of “Optuna vs Code Llama” (ICCVW 2025).
2022–2023

Software Developer and Analyst

ROSBANK, ex Société Générale Group, Moscow

Built and maintained ML models for core banking use cases; delivered analyses and reports from closed internal data; mentored interns and junior colleagues.
2021–2022

Data Analyst

Glowbyte Consulting, Moscow

Oracle PL/SQL development, Power BI dashboards, and A/B testing for top-tier banks and major food retail chains.

Education

2023–2026

M.Sc. Computer Science

Julius‑Maximilians‑Universität Würzburg, Germany

Thesis: Enabling Symbolic Reasoning with Advanced RAG Systems, written at the Bosch Center for Artificial Intelligence (BCAI) in Renningen, Germany.
2020–2022

M.Sc. Applied Mathematics and Computer Science

Lomonosov Moscow State University, Moscow, Russia

Thesis: Construction of a Disorder Predictor in Non-stationary Electroencephalography Time Series.
2016–2020

B.Sc. Applied Mathematics

National University of Science and Technology MISIS, Moscow, Russia

Thesis: Intelligent Method for Real Estate Price Prediction in Moscow Using a Machine Learning Model.

Skills

Programming

Python, SQL, Java, C++

ML / Deep Learning

PyTorch, Hugging Face Transformers, Scikit-learn, NumPy, Pandas, OpenCV

MLOps & Infra

Docker, Azure, MLflow, Kubernetes, Git

LLM & AutoML

Fine-tuning (PEFT/LoRA), Prompt Optimization (DSPy, GEPA), RAG Pipelines, Ontology-based Reasoning, Neural Architecture Search

Tools

Jira, Confluence, Power BI, Excel, LaTeX

Spoken

English (fluent), German (B1), Russian (native)