From discovery to production delivery.
I connect customer requirements, solution architecture, evaluation and cross-functional execution.
→Professional summary
Applied AI. Built for production.
Solutions engineering leader with 10+ years across engineering, analytics and machine learning. I architect agentic systems, LLM evaluation, human-in-the-loop workflows and data platforms for enterprise teams—from technical discovery and proof of concept to production delivery.
Belgrade, Serbia · CET / CEST · AI · data · product
Turn requirements into an executable system.
Define scope, interfaces, risks, evaluation criteria and delivery plans with client and engineering teams.
Build agentic systems with quality controls.
Combine tool use, observability, LLM evaluation and human feedback for dependable production behavior.
00 / Selected projects
Selected projects and outcomes.
Delivery examples spanning enterprise AI, evaluation, data operations, machine learning and product ownership.
Human-in-the-loop operations for production AI.
Designed the Human-in-the-Loop (HITL) operating layer for Tendem: expert routing, quality controls, delivery metrics and feedback loops. I own the product direction and technical architecture connecting expert operations with the core system.
Product ownerHITL architectureProduction
Visit Tendem ↗
- 01Workflow designTurning expert work into repeatable product operations.
- 02Quality systemDefining review, escalation and performance controls.
- 03Product feedbackConnecting operational signals to roadmap decisions.
Enterprise AI consulting across top-tier teams.
Led technical discovery and solution design across dozens of engagements, partnering with client engineering, product and operations teams. Scope includes proof-of-concepts, evaluation strategies, production architectures and scaled rollout.
AmazonAnthropic
JetBrainsMicrosoft
Shopify+ more frontier AI & global technology teams
22K+ product collections. Days to deliver at ≥95% accuracy.
Shopify needed high-quality labeled data despite noisy, user-generated inputs. I coordinated a 10-person cross-functional team of engineers and managers to map 22K+ product collections to a 10K+ taxonomy using retrieval-augmented generation (RAG), structural tree search and targeted human review.
22K+ product collections10K+ categories10-person teamDays-long deadline
Read public case ↗
From 400 evaluations a week to 3,000 a day.
Customized human-evaluation pipelines scaled multi-turn agent research while tracking responsiveness, transparency, accuracy, groundedness and helpfulness.
400 / week3K / day5 criteria
Read public case ↗
Stress-testing AI agents in virtual companies.
High-fidelity business environments combine domain experts, solution engineering, mocked tools and deterministic evaluation to expose weaknesses before enterprise rollout.
50+ tools40+ languages≥50% fail target
Read public method ↗
Multimodal learning for a global education platform.
Built internal data products, retrieval-augmented support and recommendation work across video, audio and text for an international education platform.
12K children*70 countries*210K lessons*
Seven-day flood forecasting for the Lena River.
Built a predictive ML solution with team Geoverflow for a real emergency-response task and received a special prize.
7-day horizon45 teamsSpecial prize
A top-15% commercial property valuation model.
Created a team ML pipeline for commercial property appraisal under hackathon constraints.
Top 15%5-person teamValuation ML
27 capstone products across 15 cohorts.
Mentored teams across computer vision, NLP, recommender systems and production ML.
15 cohorts27 projectsApplied AI
* Kidskey figures describe platform scale, not attributed project uplift.
01 / Conferences
Conference representation.
Representing teams at major NLP and Web3 events while connecting research and industry conversations to product work.
Official Toloka host
at EMNLP 2025.
Represented Toloka at the 30th Conference on Empirical Methods in Natural Language Processing, connecting research conversations with production data and evaluation practice. Toloka participated as a Platinum sponsor.
- Platinum sponsor
- Booth #1
- Suzhou, China
Paris Blockchain Week
at the Louvre.
Joined the fifth edition at the Carrousel du Louvre while working on ML systems in the crypto domain—following conversations across open finance, AI, regulation and Web3 infrastructure.
- 8–12 April
- Carrousel du Louvre
- Paris, France
02 / Skills
Technical skills.
A focused toolkit for AI architecture, machine learning, data engineering, cloud infrastructure and production delivery.
Agentic engineering · Daily practice
Production workflows for coding and multi-agent systems.
Daily use of proprietary and open-source agent harnesses, reusable skills, tool integrations, critic loops and evaluator-driven orchestration.
Proprietary harnessesClaude Code · Codex
Open-source harnessesOpenCode · Hermes
ExtensionsSkills · Plugins · MCP · Tool use
Control patternsCritic loops · Evaluator loops · Long-running loops · Multi-agent orchestration
Foundation
- Python
- Unix
- SSH
- Git
- Docker
- SQL
- Pydantic
ML & AI
- PyTorch
- Scikit-learn
- CatBoost
- LightGBM
- XGBoost
- Hugging Face
- LLM evaluation
Data systems
- Pandas
- Polars
- PySpark
- Airflow
- Kafka
- Pub/Sub
- Databricks
- ClickHouse
- Temporal
Cloud & storage
- GCP
- AWS
- Azure
- BigQuery
- PostgreSQL
- Redis
- Grafana
03 / Work experience
Work experience.
Professional roles across solutions engineering, technical leadership, machine learning, data science and environmental analytics.
2024 — presentCurrent role
Senior Solutions Engineer & Technical Team Lead
Toloka ↗Lead a Solutions Engineering team delivering enterprise AI programs across frontier models, e-commerce and developer tooling. Own technical discovery, architecture reviews, delivery planning and client communication.
- Coordinate cross-functional delivery across solutions engineers, data scientists, delivery managers and client teams.
- Translate customer requirements into proof-of-concepts, evaluation plans, production architectures and rollout decisions.
- Align engineering priorities and internal product development with customer and commercial objectives.
- Own the Human-in-the-Loop (HITL) product stream for Tendem AI ↗, combining expert operations, quality control and product strategy.
- Design and deliver expert-generated training and evaluation datasets used to develop frontier models, supporting Supervised Fine-Tuning (SFT), Reinforcement Learning from Human Feedback (RLHF) and large language model evaluation.
2023 — 2024Full-time · Cyprus
Data & ML Lead
Machine Learning Engineer
Kidskey ↗Owned the startup’s data and machine learning function, leading a cross-functional team of analysts, software engineers and ML engineers across product discovery, model prototyping, data pipelines, backend integration and stakeholder delivery for an international education platform.
2023 — 2024Contract · Switzerland
Data & ML Lead
Machine Learning Engineer
IamFutureOwned data and ML delivery for the startup, leading a cross-functional team of analysts, software engineers and ML engineers from proof of concept to MVP across crypto time series, natural language processing (NLP) and backend services.
2021 — 2023Remote
Data Scientist & Mentor
Elbrus Coding BootcampOwned analytics and ML initiatives across experimentation, marketing and a Skolkovo-backed AI project; delivered technical mentorship across the Data Science program.
2016 — 2020Multiple contracts
Engineer & Data Analyst
Gazprom Seaprojects · MSU Marine Research Center · EcoskyAutomated pipelines, analyzed remote-sensing data, built digital maps and led environmental engineering projects.
Earlier experience
Lead Environmental Consultant
EcoskyLed monitoring projects, developed analytical reports for multinational energy companies and provided GIS support.
Junior Research Scientist
Dokuchaev Soil Science InstituteAnalyzed 2,000+ soil samples during a two-month field expedition from the Black Sea to the Caspian Sea. Developed remote-sensing soil-mapping workflows with QGIS, SAGA GIS, R, SVM and Random Forest. Optimized ICP-AES laboratory data processing.
Participated in international conferences:
- Third International Salinity Forum — Riverside, USA, 2014
- Pedometrics 2015 — Córdoba, Spain
Environmental Technician
Lomonosov Moscow State UniversityConducted fieldwork on industrial impact within conservation landscapes.
04 / Education
Education.
Lomonosov Moscow State University
Specialist in Environmental and Natural Resource Management
Faculty of Geography · Landscape Geochemistry and Soil Geography
Data Science
National University of Science and Technology MISIS
Refresher training in machine learning and data analysis
Contact
Open to thoughtful conversations.
About enterprise AI, solutions engineering leadership, product ownership and complex technical delivery.