Meta Platforms
Research Scientist Intern, On-Device AI Model Optimization (PhD)
Meta Platforms, Sunnyvale, California, United States, 94087
Research Scientist Intern, On-Device Ai Model Optimization (PhD)
Reality Labs focuses on delivering Meta's vision through Virtual Reality (VR) and Augmented Reality (AR). Enabling compelling user experiences on Virtual and Augmented Reality devices requires innovation and co-design across all layers of stack from novel algorithms to custom silicon. The Meta AR/VR team is driving the state of the art forward with breakthrough work in computer vision, speech, virtual assistant, machine learning, mixed reality, graphics, displays, sensors, and new ways to map the human body among many others. We are seeking exceptional interns with a background in developing efficient models for AR/VR applications, including, but not limited to, computer vision, speech, sensor fusion, NLP, multi-modal AI. Our team focuses on developing state-of-the-art models with tight constraints on memory and power and on model optimization techniques, e.g., sparsity, quantization, compression, network architecture search, tensor decomposition, etc. The ideal candidate will have practical experience in developing real time models that can achieve high accuracy under deployment constraints such as limited compute and memory resources. In this position, you will get exposure to the full stack from user experiences, algorithms down to hardware execution blocks. You will work with the domain experts to understand the challenges and build state-of-the-art networks to tackle them and then work with the software/hardware team to deploy these solutions on-device. Our internships are twelve (12) - sixteen (24) weeks long and we have various start dates throughout the year. Responsibilities
Define, plan and execute cutting-edge deep learning research to advance AR/VR experiences Develop novel deep learning techniques to achieve state-of-the-art accuracy within the constraints of on-device and real-time execution on LLM, multimodal models and reasoning models Collaborate with other research scientists and software engineers to develop innovative deep learning techniques, such as quantization, compression, pruning etc, for vision, speech, user interface and other use-cases Collaborate with software and hardware engineers to develop tradeoff curves for accuracy vs the runtime resources/constraints such as latency, energy Communicate the experimental results and the recommendations clearly, both within the group as well as to the cross-functional groups Minimum Qualifications
Currently has, or is in the process of obtaining, a PhD degree in Computer Science, Electrical Engineering or related field Must obtain work authorization in the country of employment at the time of hire and maintain ongoing work authorization during employment Experience in developing machine learning solutions for problems Programming experience and experience with deep learning framework for prototyping ideas Interpersonal experience: cross-group and cross-culture collaboration Preferred Qualifications
Intent to return to degree-program after the completion of the internship/co-op Publication track record of first-authored publications in top-tier machine learning conferences such as NeurIPS, ICML, ICLR, ACL, AAAI and other similar venues Experience in training and inference of LLMs, multimodal models or reasoning models Experience in one or more model optimization techniques, including quantization, pruning, distillation, Low-Rank Adaptation (LoRA), Parameter-Efficient Fine-Tuning (PEFT) etc, for cloud or on-device AI deployment Demonstrated research and software engineering experience via an internship, coding competitions, or contributions in open source repositories (e.g. GitHub)
Reality Labs focuses on delivering Meta's vision through Virtual Reality (VR) and Augmented Reality (AR). Enabling compelling user experiences on Virtual and Augmented Reality devices requires innovation and co-design across all layers of stack from novel algorithms to custom silicon. The Meta AR/VR team is driving the state of the art forward with breakthrough work in computer vision, speech, virtual assistant, machine learning, mixed reality, graphics, displays, sensors, and new ways to map the human body among many others. We are seeking exceptional interns with a background in developing efficient models for AR/VR applications, including, but not limited to, computer vision, speech, sensor fusion, NLP, multi-modal AI. Our team focuses on developing state-of-the-art models with tight constraints on memory and power and on model optimization techniques, e.g., sparsity, quantization, compression, network architecture search, tensor decomposition, etc. The ideal candidate will have practical experience in developing real time models that can achieve high accuracy under deployment constraints such as limited compute and memory resources. In this position, you will get exposure to the full stack from user experiences, algorithms down to hardware execution blocks. You will work with the domain experts to understand the challenges and build state-of-the-art networks to tackle them and then work with the software/hardware team to deploy these solutions on-device. Our internships are twelve (12) - sixteen (24) weeks long and we have various start dates throughout the year. Responsibilities
Define, plan and execute cutting-edge deep learning research to advance AR/VR experiences Develop novel deep learning techniques to achieve state-of-the-art accuracy within the constraints of on-device and real-time execution on LLM, multimodal models and reasoning models Collaborate with other research scientists and software engineers to develop innovative deep learning techniques, such as quantization, compression, pruning etc, for vision, speech, user interface and other use-cases Collaborate with software and hardware engineers to develop tradeoff curves for accuracy vs the runtime resources/constraints such as latency, energy Communicate the experimental results and the recommendations clearly, both within the group as well as to the cross-functional groups Minimum Qualifications
Currently has, or is in the process of obtaining, a PhD degree in Computer Science, Electrical Engineering or related field Must obtain work authorization in the country of employment at the time of hire and maintain ongoing work authorization during employment Experience in developing machine learning solutions for problems Programming experience and experience with deep learning framework for prototyping ideas Interpersonal experience: cross-group and cross-culture collaboration Preferred Qualifications
Intent to return to degree-program after the completion of the internship/co-op Publication track record of first-authored publications in top-tier machine learning conferences such as NeurIPS, ICML, ICLR, ACL, AAAI and other similar venues Experience in training and inference of LLMs, multimodal models or reasoning models Experience in one or more model optimization techniques, including quantization, pruning, distillation, Low-Rank Adaptation (LoRA), Parameter-Efficient Fine-Tuning (PEFT) etc, for cloud or on-device AI deployment Demonstrated research and software engineering experience via an internship, coding competitions, or contributions in open source repositories (e.g. GitHub)