The project is about exploring hardware-algo co-design research. The candidate is expected to have excellent knowledge in fundamentals of deep learning, PyTorch, large language and vision language models, efficient inference and PEFT methods. Prior expertise in low level kernel optimization and computer architecture is a plus.
The candidate is expected to drive research in the broad areas of:
1.Understanding and improving the long context learning of foundation models.
2.Efficient LLM to SLM transformation.
3.Understanding and improving sub-quadratic attention for foundation models.
The candidate is expected to have prior publication records in top-tier venues including NeurIPS, ICML, ICLR, CVPR, ACL or similar conferences with demonstration of driving research capabilities.
This assignment is for the Fall semester.
This is an internship position and compensation will be given accordingly.
You must possess the below minimum qualifications to be initially considered for this position. Preferred qualifications are in addition to the minimum requirements and are considered a plus factor in identifying top candidates. Experience would be obtained through a combination of prior education level classes, and current level school classes, projects, research, and relevant previous job and/or internship experience.
Minimum Qualifications:
·The candidate must be pursuing a Master degree or PhD degree in Computer Science, Electrical Engineering or any STEM related field.
·Candidate must have at least one semester remaining at school, after completion of the internship.
Preferred Qualifications:
·Experience in AI/ML algorithm development.
·Experience in AI/ML foundation model fine-tuning or inference.
·Experience in working with popular foundation model LLM and VLM benchmarking datasets.
·In-depth knowledge of the model architecture and functional description of the transformer and sub-quadratic attention.
·Experience with various techniques to optimize models for compute/perf (quantization, compression, sparsity, pruning, low precision) and novel operation reparameterization methods (like shift-add, Flash Attention, linear-attention etc.).
·Experience with data precision, floating point vs. fixed point computing trade-offs.
·Hands on experience working with Git (PR/MR workflows with GitHub or any Gitlab repos).
·Experience in mapping between Neural Networks architectures and hardware accelerated inference.
Benefits:
We offer a total compensation package that ranks among the best in the industry. It consists of competitive pay, stock, bonuses, as well as, benefit programs which include health, retirement, and vacation. Find more information about all of our Amazing Benefits here: https://www.intel.com/content/www/us/en/jobs/benefits.html
Annual Salary Range for jobs which could be performed in
US, Colorado, New York, Washington, California:$63,000.00-$166,000.00Salary range dependent on a number of factors including location and experience.
Work Model for this Role
This role is available as a fully home-based and generally would require you to attend Intel sites only occasionally based on business need. This role may also be available as our hybrid work model which allows employees to split their time between working on-site at their assigned Intel site and off-site. In certain circumstances the work model may change to accommodate business needs.The application window for this job posting is expected to end by 08/09/2024Yearly based
Virtual - USA AZ