Job Opportunities

The codesign group at Facebook has a number of full-time job openings for research scientist, software engineer, and engineering manager roles.

About the Team

The AI SW/HW Co-Design team is a part of Facebook technology and strategy organization. The team’s mission is to explore, develop and help productize high-performance software and hardware technologies for AI at datacenter scale. The team co-optimizes both software (e.g., algorithms and numerics) and hardware (e.g., platform and network) to come up with a balanced system design. Developing new systems requires understanding performance bottlenecks on existing systems. Therefore, the team also works on optimizing performance of AI models running in production. This has resulted in TCO wins for all key AI services.

The team makes impact by working on both pathfinding and productization: it builds prototypes to demonstrate the value of the new ideas, and it works closely with many partner teams to bring them into production.

Here are some of our recent publications, if you are interested in learning more about the kind of work that we do:

• DL Models and benchmarks
  • Deep Learning Recommendation Model for Personalization and Recommendation Systems
  • Time-based Sequence Model for Personalization and Recommendation Systems
• Algorithms
  • Compositional Embeddings Using Complementary Partitions for Memory-Efficient Recommendation Systems, KDD, 2020
  • Mixed Dimension Embeddings with Application to Memory-Efficient Recommendation Systems, CoRR, 2019
• Systems
  • Deep Learning Training in Facebook Data Centers: Design of Scale-up and Scale-out Systems, CoRR, 2020
  • Deep Learning Inference in Facebook Data Centers: Characterization, Performance Optimizations and Hardware Implications, CoRR, 2018
  • Bandana: Using Non-volatile Memory for Storing Deep Learning Models, SysML, 2019
• Quantization techniques
  • On Periodic Functions as Regularizers for Quantization of Neural Networks, CoRR, 2018
  • Post-Training 4-bit Quantization on Embedding Tables, MLSys Workshop@NeurIPS, 2019
  • Training Deep Learning Recommendation Model with Quantized Collective Communications, 2020

AI System Codesign Research Scientist or Software Engineer Roles

Roles and Responsibilities:

• HPC systems software expert
• [For Software Engineer Roles] Strong programming ability: Ranging the spectrum of prototype software to production
• [For Research Scientist Roles] Proven track record of publishing in top-tier conferences
• Performance at the core: Ability to identify and improve software and hardware performance bottlenecks
• Computing-at-scale: Enabling large-scale software deployments in production
• Ability to work with and develop software solutions for cutting-edge (including off-market) hardware systems
• Ability to work across team boundaries

Qualifications/Experience/Skills Required (it is OK to have expertise in only a subset of them):

• Experience with some subset of the following HPC systems software: Accelerator (GPU/ASIC) kernel development and optimization (e.g., NVIDIA, AMD, Intel, or other misc accelerator), CPU-based threading models (e.g., OpenMP, TBB, Pthreads), HPC communication libraries (e.g., NCCL, MPI), parallel runtime software solutions, numerical libraries (e.g., mixed precision linear algebra, tensor-based frameworks), performance enablement, tracing, profiling and debugging.
• Scientific computing or other forms of HPC with an AI/ML/DL emphasis
• Extensive programming experience. Knowledge of C/C++ and Python (highly skilled in at least one of them).
• Scaling research across different modalities
• Experience with AI workload optimizations (compression, quantization, pruning techniques; graph-based systems)
• Experience with distributed AI training and inference (data-center-scale distributed training/inference is preferred)
• Performance, programmability, and efficiency at data-center scale (data access optimizations such as prefetching and caching, designing scalable frameworks for efficient use of high performance hardware, high-performance and fault-tolerant middleware, network and communication fabric optimization)

To Apply

For full-time positions, please apply here. Unfortunately, the posting is shared between multiple teams, so if you are interested in the above mentioned roles, please reach out to Pavan Balaji, Maxim Naumov or Sam Naghshineh expressing your interest and the specific kind of role that you are interested in as well.

For internship positions: Bachelors and masters students, please apply here; Ph.D. students, please apply here.