Siarhei Thor
AI/ML Engineer | Applied ML, LLM Systems & AI Product Infrastructure
Zurich, Switzerland | (+41) 78 217 35 65 | siarheithor@gmail.com
Last updated: June 15, 2026
Profile
AI/ML Engineer experienced in owning the full lifecycle of production AI systems, from data preparation, feature engineering, model training, evaluation, and calibration through API development, deployment, monitoring, and automated retraining. Builds predictive ML services, agentic LLM applications, and backend infrastructure that turn complex real-world data into reliable internal tools and user-facing products. Brings hands-on engineering, product thinking, and experience leading complex multidisciplinary projects.
Experience
AI/ML Engineer
May 2024 - Present
- Developed end-to-end ML services for running and cycling analytics, spanning data preparation, model training, evaluation, calibration, deployment, monitoring, and automated retraining workflows.
- Designed feature engineering and data-quality workflows for time-series data from sports wearables, including preprocessing, bad data detection and handling, and support for incomplete activity data.
- Built an AI-assisted pipeline for collecting, enriching, and maintaining 5,000+ structured RaceDB records, replacing a part-time manual scraping and data-maintenance workflow.
- Designed AI-assisted training-planning and personalized feedback systems combining LLMs, sports analytics, structured knowledge, and custom training-load metrics.
- Developed AI coding-assistant workflows for repository maintenance, tool orchestration, sandboxed execution, reusable agent skills, and knowledge graph-supported project knowledge bases.
Data Scientist / AI Engineer
Sep 2023 - Apr 2024 Jun 2025 - Sep 2025, project-based
- Built conversational analytics products on top of survey datasets, including chat-based insight workflows that translate structured data into natural-language answers.
- Developed a multi-agent survey intelligence prototype combining semantic retrieval, SQL-based analysis, statistical guardrails, and user-facing chat interfaces.
- Built preprocessing, vector-store, API, and interactive-analysis pipelines using FastAPI, DuckDB, pandas, SciPy, and FAISS.
AI Strategy and Digitalization Consultant
Sep 2021 - Aug 2023
- Advised architecture and real-estate clients on AI adoption, translating domain problems into practical workflows and MVP concepts.
- Built applied AI prototypes including a GPT-powered regulatory assistant for architects, a Stable Diffusion / ControlNet workflow for design exploration, and digital tools for sustainability-focused planning analysis.
Senior Architect / Computational Design
Jan 2015 - Oct 2021
- Led and contributed to large-scale building projects across Sweden and the Netherlands, coordinating complex design-development work across international teams, consultants, and clients.
- Used programming and parametric design tools to improve delivery quality, automate repetitive workflows, and support complex design coordination.
- Managed multidisciplinary teams and developed training material for junior architects in computational design and digital methods.
Technical Skills
Programming Languages
Python, SQL
ML, AI & Modeling Frameworks
scikit-learn, XGBoost, TensorFlow, PydanticAI, Pydantic, OpenAI-compatible APIs
Data & Analytics Libraries
pandas, NumPy, SciPy, DuckDB, FAISS, PyArrow, Optuna, Plotly
Application & Data Systems
FastAPI, SQLAlchemy, PostgreSQL, MySQL, Redis, WebSockets, Alembic
Infrastructure, Observability & Tooling
Docker, Gunicorn/Uvicorn, OpenTelemetry, Langfuse, Sentry, pytest, Git, Linux, GCP
Selected Projects
Knowledge Hub
Jul 2025 - Present
- Built a subject-agnostic, chat-native guided-learning engine with plugin-based content, mastery-driven quizzes, scheduled delivery, persistent learner state, and multi-provider LLM orchestration.
- Shipped German as the first subject across Telegram and web, including CEFR-based progression, cached AI-generated content, and real-time voice roleplay.
Crypto Lens
2026 - Present
- Built an always-on agentic market-intelligence system that aggregates price, technical indicators, news, sentiment, and macro data into scheduled crypto updates.
- Served a private research bot and facts-only public channel from one pipeline, enforcing the facts-versus-verdict boundary in code and suppressing low-information updates.
Education & Training
MSc, Architectural Engineering — Delft University of Technology, The Netherlands
September 2009 - June 2014
Data Science Program — EC Education, Stockholm / Online
August 2023 - June 2024
Additional
Languages: Russian: native; English, Dutch, Swedish, and Belarusian: fluent.
Interests: Orienteering, trail running, cycling, photography.