NLP Machine Learning Engineer
Yandex
Sep 2025 — PresentBuilt and productionized LLM/VLM prompt-optimization, evaluation, deployment, and monitoring pipelines; ran A/B experiments and automated quality validation.
IM
Igor Morozov
@Kisponis
Machine Learning Engineer | LLM/VLM, Computer Vision & ML Infrastructure
I’m an ML Engineer working across LLM/VLM systems, computer vision, and 3D reconstruction. I’m especially interested in applied ML research, model evaluation and optimization, and turning experimental ideas into reliable production systems.
Career
Experience: 3 yearsExperience
Machine Learning Engineer
TomTom
Aug 2024 — Sep 2025Built an end-to-end Gaussian Splatting pipeline for urban scenes over 1 TB and led migration of GPU orchestration to Kubernetes and Argo across up to 80 GPUs.
Commercial Forecasting Specialist
METRO AG
Jun 2023 — Aug 2024Owned daily demand forecasting for about 15 retail chains and reduced modified MAPE by 15 pp using ARIMA plus a custom daily-allocation optimization layer.
Education
M.Sc. in Modern Computer Science, Faculty of Applied Mathematics
National Research University Higher School of Economics (HSE)
2025 — 2027Master’s program in modern computer science with a focus on machine learning and applied research.
Data Science
Yandex School of Data Analysis
2025 — 2027Data Science program focused on machine learning, statistics, and applied research.
B.Sc. in Applied Mathematics and Informatics
National Research University MPEI
2021 — 2025Bachelor’s degree in applied mathematics and informatics.
Selected projects
LLM/VLM Prompt Optimization and Evaluation Framework
Built a reusable framework for versioned prompts, datasets, automated evaluation, experiment comparison, and error analysis. Optimized image-relevance classification with GEPA and validated gains through online A/B testing.
Large-Scale Urban Gaussian Splatting Pipeline
Built an end-to-end 3D reconstruction pipeline for scenes over 1 TB and tens of thousands of images, including COLMAP, training, fine-tuning, block merging, monitoring, and GPU orchestration.
Retail Demand Forecasting System
Developed a hybrid ARIMA and optimization-based forecasting system for about 15 retail chains, reducing modified MAPE by 15 percentage points while supporting manual delivery-calendar constraints.
Achievements
VLM classifier surpassed generalist human annotators
Yandex · 2026
Optimized a multimodal prompt with GEPA for image-relevance classification and outperformed generalist annotators against expert labels on a 2,000-query validation set.
Urban 3D reconstruction reached demo-ready quality
TomTom · 2025
The resulting urban scene was presented to Moscow city leadership and contributed to securing follow-on project funding.
Skills
PythonPyTorchLLMVLMNLPComputer VisionGaussian Splatting3D ReconstructionKubernetesArgo WorkflowsFastAPIMLOps