Skip to content
View dmitrykhursen's full-sized avatar
1 follower · 0 following

Achievements

Achievement: Pull Shark

Achievements

Achievement: Pull Shark

Highlights

  • Pro

Block or report dmitrykhursen

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
dmitrykhursen/README.md

Hi, I'm Dmytro 👋

AI/ML Engineer working on multimodal LLMs, LLM data pipelines and multi-GPU training. 3+ years of automotive R&D at Valeo, including 2.5 years at valeo.ai fine-tuning vision-language models on a supercomputer. MSc in Artificial Intelligence, CTU Prague (2026).

📍 Prague · 🇺🇦 Ukrainian · open to AI/ML engineering roles

Featured work

VQA for Autonomous Driving · MSc thesis, CTU × valeo.ai Fine-tuned InternVL2-2B with LoRA on 8× A100 for driving-scene question answering. A distribution-aligned 25k-question split beat 12× more data, LLM-generated pseudo-labels pushed the score further, and an oracle study showed visual grounding is the bottleneck. Demo · Models & dataset · Thesis

Mu-SHROOM: LLM hallucination detection · team lead, team of 6 Span-level hallucination detection (SemEval-2025 Task 3 setup): Wikipedia retrieval with Sentence-BERT + FAISS feeding a Gemma-3-27B LLM-as-a-judge.

Freshlabels brand-equity case · EY-Parthenon Challenge 2026, top 5 of 50 Python scrapers, fuzzy product matching and Streamlit dashboards for a consulting case, built with AI coding agents.

Hands-Free: road-scene perception · University of Vermont, 1st place UVM CS Fair 2022 Traffic-sign CNN on top of Faster R-CNN detections.

Tools I use most

Python · PyTorch · Hugging Face Transformers & PEFT · LoRA / QLoRA · DDP / FSDP · SLURM / HPC · vLLM · FAISS · YOLO · SAM · OpenCV · Git · Linux · Claude Code

Contact

LinkedIn · Hugging Face · dmytro.khursenko@gmail.com

Pinned Loading

  1. mushroom mushroom Public

    Project on SemEval-2025 Task-3 - Mu-SHROOM, the Multilingual Shared-task on Hallucinations and Related Observable Overgeneration Mistakes

    Jupyter Notebook 2

  2. VQA-AD-CTU VQA-AD-CTU Public

    Fine-tuning Vision-Language Models for autonomous driving VQA: custom i.i.d. DriveLM-nuScenes split, ~132k pseudo-labels generated by Qwen3 from nuScenes sensor priors (3D bounding boxes, LiDAR dep…

    Python 2

  3. eyxfreshlabels eyxfreshlabels Public

    EY-Parthenon Challenge 2026 (top 5 of 50): brand-equity case for Freshlabels; Python scrapers, fuzzy product matching, Streamlit dashboards.

    Python 1

  4. hands-free-driving-vermont hands-free-driving-vermont Public

    Road-scene analysis: Faster R-CNN + traffic-sign CNN + lane detection (UVM CS Fair 2022, 1st place)

    Python 1