We have several openings for students projects :
1. Microfluidics-guided personalized cancer immunotherapy
We have previously established a microfluidic platform for testing therapy options directly on tumor biopsies from patients (Eduati F et al., Nature Communications 2018, Utharala R et al., Nature Protocols 2022, Mathur L et al., Nature Communications 2022). However, these platforms are not yet suited for predicting outcomes of immune therapies. The aim of this project is to establish different kinds of assays for assessing the efficient killing of cancer cells by T-cells. The primary readout will be based on light-sheet imaging. The successful candidate will optimize staining of different cell types using fluorescent antibodies and establish functional fluorescence assays. Prior experience with cell culture, imaging and/or image analysis is a strong plus.
Keywords: cancer immune therapy, microfluidics, imaging
Supervisor: Christoph Merten
Co-supervisor: Tianhao Li
Contact: [email protected], [email protected]
2. Microfluidic single cell screening of tumor-reactive T-cells
We have previously established droplet microfluidic platforms for screening antibodies at the single cell level (El Debs B et al., PNAS 2012, Shembekar N et al., Cell Reports 2018, Panwar J et al., Nature Protocols 2023). In these systems, tiny aqueous droplets surrounded by oil serve as miniaturized test tubes. The technology enables the screening of hundred thousands of antibodies in a single experiment and has led to the establishment of a startup company in 2017 (Veraxa). The aim of this project is the establishment of similar screening platforms for T-cell therapies. The successful candidate will establish single cell droplet microfluidic assays using patient-derived model systems. Prior experience with cell culture, PCR and/or NGS is a strong plus.
Keywords: T-cell therapy, single cell analysis, microfluidics
Supervisor: Christoph Merten
Co-supervisor: Roger Diaz Codina
Contact: [email protected], [email protected]
Required: Full time master thesis
3. Master thesis position:
Bulk, chemical-free demulsification of water-in-oil microfluidic droplets by electromagnetic manipulation
Water-in-oil (W/O) microfluidic droplets are extensively used in single-cell genomics/transcriptomics, directed evolution, and diagnostic assays. A crucial step following droplet-based incubation or reaction is the recovery of the aqueous contents (e.g., cells, nucleic acids, proteins) for downstream analysis. Current methods predominantly rely on chemical surfactants to break the emulsion. However, these chemicals can interfere with sensitive molecular biology assays (like PCR or enzymatic reactions) and may affect the viability or integrity of recovered cells, limiting the scope and accuracy of subsequent analyses. Therefore, a gentle, rapid, and chemical-free method for droplet demulsification is highly desirable.
We propose to develop and validate a novel approach for the bulk, chemical-free demulsification of W/O droplets using electromagnetic fields. This project aims to design, build, and test an electrical apparatus capable of inducing controlled droplet coalescence in a dedicated module, enabling efficient recovery of droplet contents without chemical additives. The system will exploit the dielectric properties of the oil and water phases to generate forces that destabilize the droplet interface upon application of specific electric fields. The project will involve investigating optimal electrode geometries and electrical parameters (voltage, frequency, waveform) to maximize coalescence efficiency while ensuring the integrity of biological samples potentially contained within the droplets.
Expertise Gained:
The project will give the student the opportunity to gain expertise in:
- Microfluidic device design and fabrication/prototyping
- Electronic system design and integration for device actuation
- Experimental methods in microfluidics (droplet generation, manipulation)
- Microscopy and image analysis
- Basic cell handling and molecular biology techniques for validation
Type of work: 10% literature study, 25% design, simulation, and prototyping, 40% experimental work and device testing, 25% data analysis and results reporting.
Duration: 4 – 6 months
Prerequisites:
- Ideally someone with a background in Electrical Engineering, Physics, Microengineering, Mechanical Engineering, or Bioengineering with relevant hands-on experience.
- Familiarity with prototyping, electronics, laboratory instrumentation (power supplies, oscilloscopes), and CAD software.
- A strong motivation and a hands-on, problem-solving attitude are required regardless of the specific background.
Do not hesitate to contact us by email in case of interest: [email protected], [email protected]
4 · Light-Sheet Microscopy Live Wire: Real-Time Capture & Cluster-Scale Processing Pipeline(Master/Bachelor/Semester Thesis)
Short Abstract
Build a “zero-latency” data path that streams raw light-sheet stacks from our pyMMCore+ acquisition GUI directly into the EPFL HPC cluster, bypassing local disks and triggering automatic denoising / segmentation jobs. The student will redesign the GUI with a stable backend, implement live detection + OME-Zarr chunking, and benchmark throughput in a long acquisition run.
Key Words
Software hardware integration · Real-time streaming · HPC workflow · Python · pyMMCore+ · Light-sheet microscopy
Description
Modern light-sheet experiments can generate almost 1 GB s⁻¹; writing to local SSDs then copying to the cluster is the current bottleneck.
Your mission is to create a fully streaming acquisition/processing pipeline:
- GUI refactor – Improve the current light-sheet microscope control software based on python. Extend the existing pyMMCore+ control panel to a producer-consumer pattern using Qt signals & slots; camera frames land in a bounded queue while a background worker forwards them over.
- On-the-fly preprocessing – integrate a lightweight bead-based focus detector and write data directly to cloud-optimized OME-Zarr chunks.
- Cluster hand-off – when a new Zarr folder appears on the NAS3 mount, launch jobs for denoising and Cellpose segmentation.
- Benchmark – Run real acquisition to benchmark sustained throughput (MB s⁻¹) and end-to-end latency. Profile CPU/GPU utilization.
- Robustness – apply standard producer–consumer stress tests to verify loss-free acquisition beyond 60 min runs.
Skills you will gain
Python image processing/ machine learning pipeline, soft-hardware integration, hands-on light-sheet microscopy.
Goal
Remove the disk bottleneck and enable real-time analytics for high-throughput droplet-LSFM drug screens.
Contact Details
If you are interested, please send your CV and transcript of records to Tianhao Li, [email protected]
5 · Cell Segmentation and Classification Pipeline of Light-Sheet Imaging Data for Personalized Cancer Therapy(Master/Bachelor/Semester Thesis)
Short Abstract
Benchmark and integrate the most effective 3-D deep-learning pipelines for nucleus + membrane segmentation in light-sheet volumes, then build a lightweight classifier that labels each segmented object by cell type and treatment status. The outcome is a fully reproducible Jupyter Book, complete with trained weights and evaluation scripts, ready to drop into the droplet-LSFM screening workflow.
Key Words
3-D segmentation · StarDist 3D · Cellpose 3D · CellSeg3D · 3-D U-Net · napari annotation · self-supervised learning · Light-sheet microscopy
Description
Accurate 3-D cell masks are the foundation of every downstream morphology metric.
Your mission is to find, document, and package the best segmentation + classification stack for our multicolor LSFM data:
- Ground-truth creation — annotate cells in napari using nuclei + membrane channels as guidance.
- Model zoo benchmark — train and evaluate StarDist 3D, Cellpose 2.0 3D, PlantSeg3D, and a lightweight 3-D U-Net on identical splits; measure mean IoU and inference speed.
- Self-/semi-supervised add-ons — test at least one self-supervised approach to assess robustness when manual labels are scarce.
- Cell-type classifier — extract shape + intensity features and compare supervised versus self-supervised embeddings for live/dead cell labels.
- Packaging — wrap the final pipeline in a Jupyter Book tutorial; export PyTorch weights and a template for cluster execution.
Skills you will gain
Machine learning image analysis pipeline, and hands-on light-sheet microscopy.
Goal
Deliver a one-click 3-D segmentation + cell-type labeling module that slots into the droplet-LSFM analysis workflow.
Contact Details
If you are interested, please send your CV and transcript of records to Tianhao Li, [email protected]