Marcin Abram

Teaching and mentoring

I have taught very different rooms: specialized graduate machine learning classes, an introductory physics lecture for 120 students, IBM engineers in internal training sessions, and client teams in workshops, both as part of enablement programs and as exploratory sessions to support go-to-market activities. I have also supervised PhD and master's students, and mentored interns and engineers.

Students

University of Southern California, 2020 to 2024

At USC I founded the Machine Learning for the Natural Sciences group. Each semester I offered research-oriented internships for students from physics, computer science, and the data science program. In numbers:

  • 30+

    students did research with me, each for at least a semester.

  • 3

    PhD students mentored to completion (three successfully defended theses).

  • 8

    capstone projects supervised.

  • 4 papers

    with students as co-authors: quantum emulators, federated learning, and two on language models and context (2023, 2024).

  • 3 awards

    to student teams in USC's DataFirst program: Best Project Achievement in two semesters and Best Data Science Insight in one. Another team placed second in the RMDS Lab data science competition.

  • Alumni

    now work at NVIDIA, Meta, IBM, Intel, and Verisk.

At IBM I supervised four interns working on applications and evaluation of language models, and co-supervised an MSc thesis on context-window limits and the evaluation of generated text.

Courses

  • 2025

    Explainable AI. A lecture series at Jagiellonian University.

  • 2024–2025

    Workshops for engineers at IBM. Information retrieval systems, language models and semantic search, and large-scale document processing. Four multi-day workshops for one of IBM's larger government clients (63 contact hours in total).

  • 2020–2024

    Physics at USC. Classical Mechanics (three times), Thermodynamics and Statistical Mechanics, Electricity and Magnetism, Electrodynamics, and Fundamentals of Physics.

  • 2021

    Machine Learning for Data Science. Graduate course at USC. Assignments involved building solutions on real data instead of textbook exercises.

  • 2013–2015

    Physics at Jagiellonian University. Statistical Physics, Physics with Elements of Biophysics, Physics Laboratory. Student evaluations of 4.88 and 4.97 out of 5, against university averages of 4.40 and 4.36.

  • 2012–2014

    Physics classes for gifted high-school students. Main organizer. Several students reached the national olympiad finals, and one won a silver medal at the International Olympiad on Astronomy and Astrophysics in 2013.

  • 2012–2014

    Mathematics tutor at a center for blind and visually impaired students.

Curriculum and reading

At USC I designed the full curriculum for a new graduate program in data science that combines physics with machine learning.

For students early in their studies I ran directed reading: a weekly meeting, one to one or in a small group, to work through assigned material and choose what to read next.