We have an open call for HOMER and xLungs projects for following positions. If you are interested in any of them please send your CV and Motivation Letter to przemyslaw.biecek at pw.edu.pl. We reserve the right to contact only selected candidates.
We are seeking highly motivated and talented researchers at different career levels to join MI2.AI research team in the exciting field of explainability and red-teaming of AI models.
There is a common recruitment policy for all positions.
We are looking for
- people interested in research who want to create methods and tools that have a real impact (not yet another paper).
- proactive people with initiative.
- people with proficiency in programming languages commonly used in AI research, at least one out of Python, R or Julia.
- solid understanding of AI/ML models, including their architecture and training processes.
- good communication skills and proficiency in academic writing; excellent problem-solving skills.
- the opportunity to learn from experienced and motivated researchers at different career levels as well as the opportunity to work with a rich network of international and interdisciplinary collaborators (for example Harvard Medical University, Vienna University, LMU, NTU, KDD Lab, Bari University and many others).
- a possibility for a tenure-track position after the completion of the postdoc.
- possibility to present research results at international conferences.
- access to our supercomputers, among them Eden with 32x A100 cards and 1.5TB of storage, a perfect place for training large DNN models.
Salary depend on experience and type of project (applied projects tend to have higher salaries than typical research projects).
Average rates range from 50-60 PLN/hour (equivalent to 8-9.5k FTE) for a student with experience, through to 80-85 PLN/hour (equivalent to 12-13k FTE) for an experienced PhD student. The base salary for a post-doc (no obligatory teaching duties) is 120-140k gross per year (this is a base salary, extras depend on activity).
Ex 1: Any object can be embedded with an appropriately trained neural network. But can objects be embedded using tree-based models? This project (an ideal candidate for MSc thesis) is about exploring different embedding construction techniques based on random forest or boosting methods.
Ex 2: Analysis of champion-challenger model pairs on subsets of observations obtained using data segmentation. The project consists of an extension of the EPP method (Elo for predictive models) based on segmented data (e.g. by k-means or by random splits). In addition to the performance analysis, a PD profile comparison analysis is also performed on the splits.
Ex 1: Development of techniques for explaining segmentation models for 1d (feature-producing models, e.g. survival models), 2d (image-producing models) or 3d data. (CT, hyperspectral images) Experience and knowledge of classical explanatory techniques (SHAP, LRP, CRP) are desirable.
Ex 2: Development of red-teaming techniques for LLaMA 2-type language models. Development of test suites testing the vulnerability of the model to a set of prompting-based attacks.
In addition to the research tasks listed above, responsibilities include mentoring junior staff (students, PhD students) and leading small research groups.
Additional information about costs of living in Warsaw
From FENS: Warsaw is the capital of Poland, so in principle it is one of the most expensive cities in the country but living in Warsaw is still much cheaper than in the majority of European cities. In general, a single person in Warsaw is able to cover all necessary expenses (flat, media, transport, food) with ca. 5000-5500 PLN per month. A typical cost of renting a 2-room flat (30-40 square meters) varies between 2000-3000 PLN, depending on the part of the city. Electricity and media (TV, Internet) are usually not included in the rent and have to be paid separately. The cost of the lunch in the vicinity of the main campus of WUT is ca. 25-40 PLN, while a monthly ticket for public transport is 110 or 180 PLN depending on the area of validity. Public health insurance is included in the post (this includes also admission to hospitals), however the medicines are paid separately.
In addition to research positions, we also offer positions for skilled programmers.
- Background in Computer Science, Mathematics, Statistics or similar.
- Experience in Computer Vision or NLP models
Scope of work:
- Training of machine learning models for tabular and image data
- Interpretable solutions for tabular and image data
- Responsible ML solutions for the healthcare domain