2006 — 2011
Bachelor of Marine Navigation
TalTech Estonian Maritime Academy.
About
I am Sergei. I like AI-assisted coding and experimenting with LLMs.
I identify myself as an LLM engineer and AI practitioner. I build my own pet projects and applied AI systems — RAG pipelines, agent workflows, LLM-powered products. I am working to build skills sufficient to enter the tech job market, in Estonia first and the EU in general.
Digital twin
I built a digital twin of myself. Ask it anything.
It knows everything about my background, interests, values, education and projects. When it cannot answer, it passes the question to me — and I can join the chat myself. It answers in English, Estonian or Russian.
Open the twin→Work history
Navigation Officer
After graduating from the Maritime Academy I worked as a navigation officer for several years.
Independent Quantitative Analyst · Sports Markets
Self-directed, for many years, mainly tennis. The work: feature engineering from match data, probabilistic decision-making with real money on the line, continuous model evaluation under uncertainty. The skills that came out of it — Bayesian reasoning, risk management, checking whether a model is actually right or just looks right.
End of 2024
A break
I decided to take a break and reconsider my views on life, because I no longer saw meaning or value in what I was doing.
Now
LLM Engineer · AI Practitioner
Building pet projects and applied AI systems — RAG pipelines, agent workflows, LLM-powered products.
Projects
Aug 2026
Autonomous Trading Floor
An equity trading simulation where four autonomous LLM agents manage virtual portfolios: researching the market, executing trades, and rewriting their own strategies based on how their past trades performed. Built on the Model Context Protocol and the OpenAI Agents SDK, with a FastAPI backend and a Vite/TypeScript dashboard.
Jun — Jul 2026
Tennis Match Research Dashboard
LLM agent engineering on a real domain. The dashboard renders four independent signals side by side for every upcoming ATP / WTA tour-level singles match — market consensus odds, a trained LightGBM probability, a surface-Elo baseline, an LLM-discovered news block — plus a deterministic “why model differs” panel whenever the model-vs-market gap exceeds 10pp. The purpose of the project is to demonstrate end-to-end ML + LLM engineering: data ingestion, feature engineering, model training, evaluation, LLM tool-calling integration, deployable interface.
Not a betting toolAug 2026
Mar 2026
Top skills
AI
Agentic engineering
Analysis
Languages
Education
Mar 2026
AI Coder: Vibe Coder to Agentic Engineer
Udemy · Ed Donner
Oct 2025 — Feb 2026
AI Engineer Core Track: LLM Engineering, RAG, QLoRA, Agents
Udemy · Ed Donner · certification
Nov 2025
Python Data Science: Data Prep & EDA with Python
Udemy · Alice Zhao, Maven Analytics
May — Oct 2025
Google Data Analytics Professional Certificate
Coursera · certification
Jun 2025
Practical Python Course: from Basics to Advanced
Wexler.io · Egor Wexler
2006 — 2011
Bachelor of Marine Navigation
TalTech Estonian Maritime Academy
Outside work
20 years of almost daily training — mainly in the gym, but also functional training like HIIT and CrossFit, or just walking in the forest. No smoking, no alcohol, healthy diet. I don't eat anything sweet other than fruit.
- Training
- 20 years, almost daily. Gym, HIIT, CrossFit.
- Diet
- No smoking, no alcohol, nothing sweet other than fruit.
- Time off
- Walking alone in nature. Meditation.
- Attention
- Strict information hygiene — I consider time and attention my most valuable resources.
- Interested in
- Fitness and sport, data analytics, neurobiology, Stoicism.
- School
- Straight-A student. Not sure if that counts as a good thing.
- Film
- Interstellar, by Christopher Nolan.
- Sportsman
- Andy Murray.
- Role model
- Pavel Durov.