Supervisor: 

Svetlana Illarionova, Head of Research Group; 

Maxim Sharaev, Head of BIMAI-Lab

Description: 

The project focuses on the automatic analysis of histological patch images of diffuse large B-cell lymphoma (DLBCL) biopsies. The tasks span three directions: binary semantic segmentation of macrophages (with high class imbalance), classification of the immune microenvironment, and multitask learning (MTL) that combines both tasks into a single pipeline.

-- What is the project about? What tasks does it involve? What experience will the intern gain?
-- Hands-on experience with pathology foundation models (UNI2, CONCH, Virchow2, CellViT)
-- Parameter-efficient fine-tuning techniques: LoRA, progressive unfreezing
-- Working with imbalanced medical datasets and specialized loss functions (Tversky, Focal)
-- Designing MTL architectures

Internship duration: 2 months 

Internship start date: flexible 

Requirements for the candidate:

Required:
Python — confident level
PyTorch — basic level (training, datasets, DataLoader)
Knowledge of convolutional neural network basics and encoder-decoder architectures
Understanding of segmentation quality metrics: Dice, IoU, Precision/Recall

Preferred:
Familiarity with Vision Transformers (ViT) and the attention mechanism
Understanding of transfer learning and fine-tuning

Plus:
Familiarity with LoRA / PEFT methods
Experience with albumentations, segmentation-models-pytorch, timm

Monthly compensation: not available

Contact person: Illarionova Svetlana, s.illarionova@skoltech.ru