Researcher - Machine Learning and Quantization

Zama is hiring!


Zama's mission is to bring end-to-end encryption to AI. Using their homomorphic development framework, companies can process their customer's data without seeing it, thereby preventing data breaches and surveillance.

Zama's solution is based on a breakthrough in homomorphic encryption, which enables doing data science and machine learning on encrypted data. Zama is open-source by design, as they believe privacy-enabling technologies should benefit the widest possible community of developers and researchers.

Zama's cofounders are Dr Pascal Paillier, one of the most renowned cryptography researcher, and Dr Rand Hindi, a serial entrepreneur who formerly founded Snips.

Job Description

The Homomorphic Development Framework team is building an open source framework for homomorphic machine learning. More precisely, we build a set of tools to help our user to train or modify their existing neural network, such that the obtained neural network is friendly with fully homomorphic encryption (FHE). Thus, with our tools, the user can develop neural networks which can be used in a context where user's privacy is ensured end-to-end, even if executed on an insecure server.

The goal of our tools is to be as easy and user friendly as possible, while reducing the accuracy of networks as little as possible. Notably, we want to not require the user to understand anything about cryptography, or, said in another way, we want to provide tools which look similar to what data scientist are already using, day to day.

Work in our team is a clever mix between research and product: on one side, we need to deeply understand and even make the research progress in some of the advanced machine learning topics, in order to be able to both have FHE-friendlyness and excellent accuracy; on the other side, we need to have a clean, strong and updatable product suite, notably due to Zama's open-source policy and willingness to accept external contributions.

We are looking for a researcher in ML, ideally with a specialization in quantized / discretized / binary networks, or at least with the will and interest to work on these subjects. To follow some FHE constraints and have networks which are as efficient as possible when running over encrypted data, we need to understand the state of the art in these subjects, and then to push the barriers even further. Notably, the position will include publishing, promoting our techniques, patenting and prototyping. Ideally, the researcher would also be interested by the engineering side of the company, and thus, participate to the development effort. Experience with ML frameworks would be highly appreciated. Experience in backends (e.g., onnxruntime) would also be a plus. Understanding of cryptography is a plus, but not required.

Your team (and thus you) will be responsible for:

  • using/understanding/creating the best techniques to train/modify neural networks which are FHE-friendly
  • prototyping ideas (and ideally developping the product with the rest of the engineering teams)
  • publishing papers in research conferences
  • patenting
  • diffusing the knowledge to the rest of the engineering team, and to Zama's users in the form of blogs or seminars organized by the company

Preferred Experience

  • experience in ML research (ideally with a Ph.D.), and ideally in discretization / quantization / binary neural networks
  • experience in ML frameworks
  • optionally: development of prototypes / production tools in python
  • optionally: experience with backends (e.g., onnxruntime)
  • optionally: experience with FHE or cryptography
  • be passionate about privacy and open source

Recruitment Process

Our process is described in detail here:

Additional Information

  • Contract Type: Full-Time
  • Location: Paris, France (75002)
  • Education Level: PhD and more
  • Possible full remote