Two years from mathematical foundations to a defended research or production outcome
The program leads from applied mathematics and classical ML through the modern stack (transformers, LLMs, agents, MLOps) to specialization and independent research. Four semesters, each combining core courses, an elective, and a continuous practicum.
Mathematical foundations: vector spaces, linear maps, and the geometry of data
Probabilistic reasoning and statistical inference as the foundation for sound conclusions in ML
Algorithmic thinking and efficient data structures for scalable ML systems
From linear models to neural networks: classical ML and first steps into deep learning
An elective course selected from the course bank
A first research experience: reproducing a published paper from start to finish
Optimization as the engine of learning: from theory to modern adaptive methods
Transformers, large language models, and agentic systems — the core stack of modern NLP
Production ML: data pipelines, reproducibility, deployment, and monitoring
Reliability, reproducibility, and honest evaluation of ML systems
An elective course selected from the course bank
An end-to-end ML system: from raw data to production serving
Multimodal AI: computer vision or audio/speech processing, chosen by the student
The AI frontier: Bayesian ML, causal inference, reinforcement learning, and graph neural networks
High-performance ML systems: GPU, distributed training, compression, and LLM serving
AI in real-world domains: science, robotics, healthcare, and product strategy
An elective course selected from the course bank
A team industrial project: production ML with real SLAs, latency, and quality constraints
Weekly discussion of frontier papers and development of your own research voice
Reliability, safety, and alignment of AI systems as an engineering discipline
A contribution to an open-source or production system: from concept to a merged PR or a deployed product
Structured guidance through writing the thesis, from problem definition to defense
Pre-thesis practicum: the final experiment and results collection for the thesis
Master's thesis: an original contribution to AI research or applied development
A consolidated framework of program outcomes — a draft assembled from individual course outcomes, still being refined by the working group.
The core team built the Yandex School of Data Analysis (ShAD), Yandex.Textbook, and olympiad training programs; mentors include practicing engineers at Google Zurich and industry research engineers. The faculty roster is still being finalized. confirmed in discussion.
Staff Software Engineer at Google Zurich, IOI 2004 gold medalist, ICPC 2007 silver medalist, co-founder of the "Pelican" olympiad training program.
A legendary competitive programmer (record peak TopCoder rating), Software Engineer at Google Zurich, champion of IOI, ICPC, Google Code Jam, and Facebook Hacker Cup.
Chief Data Scientist at Juno AI, PhD student at HSE; Graph ML, NLP, and LLM agents, publications at NeurIPS/RecSys (identification preliminary).
Additional courses that fill the elective slots in each semester. Rule (draft): one elective slot per semester in semesters 1–3; selection takes place at the start of the semester, a track launches once the minimum group size is reached, with priority given by order of application.
Honest status: some parameters are still under discussion by the working group — such items are marked "to be announced," not hidden.