Privacy-Preserving Engineer - Q

Описание

Our team works at the intersection of machine learning, hardware, and software, creating technologies that reshape how people communicate and experience the world around them.

Responsibilities

• Translate advanced mathematical theory, privacy mechanisms, and research formulations into robust, practical machine learning systems
• Dive deep into core model training loops (in PyTorch) to modify, inject, and optimize custom components, loss functions, clipping mechanisms, and private optimizers
• Design, architect, and implement end-to-end privacy-preserving machine learning pipelines
• Evaluate privacy-utility trade-offs, measuring model accuracy, convergence behavior, and empirical privacy leakage
• Collaborate closely with theoretical research scientists to validate algorithmic hypotheses and verify theoretical guarantees in practical setups
• Package, document, and present research implementations, technical prototypes, and proof-of-concepts clearly to cross-functional teams and leadership
• Stay updated on modern privacy-enhancing technologies and open-source differential privacy tooling

Requirements

• M.Sc. in Computer Science, Electrical Engineering, Data Science, Applied Mathematics, or a related technical field
• 3+ years of hands-on experience developing, designing, and training machine learning and deep learning models
• Strong software engineering foundation with advanced proficiency in Python and PyTorch, including deep familiarity with modifying internal training loops, custom optimizers, and gradient computation
• Solid mathematical maturity with the ability to understand theoretical papers, algorithmic proofs, and privacy mechanisms from first principles
• Demonstrated experience in privacy-preserving technologies (e.g., Differential Privacy, private aggregation, or privacy-budget accounting)
• Strong technical presentation and communication skills in English, with the ability to effectively showcase technical prototypes to diverse audience

Nice to have

• Hands-on experience with open-source differential privacy libraries (e.g., Opacus, OpenDP, Diffprivlib, or TensorFlow Privacy)
• Experience implementing DP-SGD, gradient clipping, noise addition mechanisms, or Renyi DP accounting in deep learning frameworks
• Exposure to cryptographic privacy primitives (e.g., Secure Multi-Party Computation / MPC, Federated Learning, or trusted execution environments)
• Experience balancing and optimizing privacy budgets against task performance on complex datasets