About
💡 Senior AI Research Engineer with 8+ years of experience across LLMs, multimodal and generative AI, computer vision, and production ML.
🎓 PhD in Computer Science with peer-reviewed Q1 publications and 130+ Google Scholar citations. Experienced in bridging research and engineering.
🚀 Background includes work at AI startups, research labs, and industry teams, as well as earlier software engineering experience with team leadership.
Core Expertise
LLMs & Multimodal: multimodal learning, VLMs, reasoning models, test-time compute, pre-training, SFT
Post-training & RL: GRPO, RLVR, RLHF, DPO, ORPO, reward modeling
Generative AI: diffusion models (SDXL, FLUX, Wan), image and video generation, ControlNet, IP-Adapter, LoRA
Distributed training: DeepSpeed (ZeRO-3), Megatron, FSDP, mixed precision
Efficient AI: distillation, quantization, inference optimization, vLLM, SGLang, NVIDIA Triton
Data & evaluation: synthetic data generation, LLM evaluation, LLM-as-a-judge
Tools: Python, PyTorch, Hugging Face, Docker, Linux, Git, SQL
Experience
- Post-trained Qwen LLMs and VLMs (SFT, LoRA, ORPO, DPO) to detect and explain quality issues in automotive requirements, with a ~30% relative improvement.
- Improved model reasoning with reinforcement learning with verifiable rewards (GRPO, RLVR), synthetic data generation, and LLM-as-a-judge evaluation.
- Optimized and served models in production with quantization, GPU profiling, vLLM, and SGLang; redesigned OCR and document processing for 5× faster parsing.
- Continued pre-training of a vision-language model for image and video captioning, video understanding, and other downstream tasks, with distributed training (Megatron, DeepSpeed ZeRO-3, FSDP, mixed precision).
- Built few-shot identity-preserving image generation and text-to-video models (SDXL, FLUX, ControlNet, IP-Adapter, Wan, LoRA).
- Distilled VLM-generated synthetic labels into a lightweight multimodal tagger for content search and recommendation; with quantization and NVIDIA Triton, this cut latency by 74%.
- Designed hyperbolic neural networks and transformer architectures that embed visual features in hyperbolic space, improving the interpretability of learned representations.
- Developed aesthetic preference learning: user-adaptive multimodal models that learn an individual's taste from a few ratings.
- Fine-tuned diffusion models (LoRA) to generate images that match a user's aesthetic preferences.
- Completed postdoctoral research on personalized aesthetic preference learning, published in peer-reviewed journals.
- Advanced the state of the art in facial attractiveness and aesthetic preference prediction with few-shot and meta-learning methods; published in Q1 journals.
- Researched recommender systems that use visual embeddings to address the cold-start problem.
- Created and released MEBeauty, a multi-ethnic dataset and benchmark with human preference ratings.
- Built generative models (diffusion, GAN, VAE) for face editing, image enhancement, and inpainting with PyTorch.
- Senior Software Engineer at Schneider Group: supervised and mentored developers and managed the full SDLC.
- Software Engineer at Yandex: software development and maintenance, SQL database design and administration.
Education
PhD, Computer Science — East China University of Science and Technology, 2017–2022
MSc, Computer Science — Shanghai Jiao Tong University, 2013–2017
BEng, Computer Science & Engineering — National Research Nuclear University MEPhI, 2005–2009
Selected Publications
Full list on Google Scholar.
Languages
English (full professional) · Russian (native) · Chinese (working knowledge)
Projects
A multi-ethnic facial beauty dataset collected in the wild, with human attractiveness ratings, first published in 2022 and now fully open and reproducible.
- Dataset: about 2,500 faces from six ethnic groups, of different genders and ages, cleaned, with person-level splits and an improved label.
- Benchmark: 21 methods, from 2006 to today's foundation models, including my own approach, all under one protocol.
- Reference model: a DINOv2-based model, ready to use in two lines of Python, with its accuracy across ethnic groups reported openly.
- Publication: the original peer-reviewed paper, "MEBeauty: A Multi-Ethnic Facial Beauty Dataset in the Wild", Neural Computing & Applications, 2022.
For non-commercial academic research only.
News
- Oct 2026Released the updated MEBeauty dataset on Hugging Face, together with an open benchmark of 21 methods and a reference DINOv2 model.
- 2026Serving as a peer reviewer for NeurIPS and ECCV.
- Mar 2026Joined Raiqon as a Machine Learning Engineer, working on LLM and VLM post-training and reinforcement learning.
- 2025Paper published in the Journal of Visual Communication and Image Representation: GAN Semantics for Personalized Facial Beauty Synthesis and Enhancement.
Contact
Open to research and engineering roles in LLM post-training, multimodal AI, and efficient AI, and to research collaborations.
dr.irina.lebedeva@gmail.com