Qubit Pharmaceuticals is a French-American deeptech startup, focusing on discovering novel molecules for complex targets in healthcare and materials science. We leverage proprietary molecular simulation and quantum-physics based modeling technology to develop our own discovery programs that we either co-develop with or license out to pharmaceutical and industrial partners. This enables us to design more effective and safer drug candidates, aiming to significantly reduce the time and investment needed for discovery. In just 18 months, Qubit Pharmaceuticals grew its portfolio to 7 programs in oncology, immunology & antivirals.
Our cutting-edge technology is based on over 30 years of research by our academic founders, and relies on three main components: in-depth expertise in computational science and high-performance computing (HPC), quantum chemistry and artificial intelligence algorithms, and a powerful, automated cloud platform for molecular simulation with chemical accuracy. We recently announced the launch of the world's most powerful AI foundation model for molecular simulation.
Qubit Pharmaceuticals is looking for an Intern (M/F) in Parallel Tensor train Arithmetics.
The perfect fit for this job
We are looking for a detail-oriented engineer who is willing to work on :
- Tensor networks and their applications in quantum simulations.
- Numerical decomposition methods and optimization challenges in high-performance computing.
- Efficient parallel routines within our Hyperion quantum emulator.
Your role
This project aims to optimize the process of approximating high-dimensional tensors (multidimensional arrays) using the Tensor Train (TT) representation. The TT decomposition breaks down complex tensors into a sequence of smaller, more manageable components, enabling highly efficient storage and computation. A key operation within the TT format is rounding, which involves truncating the TT-ranks while ensuring the approximation error remains within a predefined bound. This truncation is typically performed using QR orthogonalization and Singular Value Decomposition (SVD) methods. The goal of this project is to propose and develop an efficient, parallel algorithm for TT rounding.
Internship Objectives During the internship, the candidate will:
- Explore and analyze state-of-the-art methods for parallel tensor train arithmetic.
- Identify the most suitable method, or propose a new one for integration into our Hyperion emulator, balancing performance and scalability.
- Implement the selected approach.
Your qualifications and skills
- Master 2 or final year of engineering school
- Strong foundation in linear algebra and a willingness to explore tensor algebra
- Programming experience in Python and programming experience in C++ and CUDA, or a strong willingness to learn
- Interest in numerical analysis and algorithm optimization.
The proposal benefits & perks
- Duration: 6 months
- Starting date: from February 2026
- Lunch vouchers worth €9 covered at 50%.
- Location: Paris 14th arrondissement
