Graduate RA Position on Error Detection for GPUs using ML

PI: Saurabh Bagchi

Posted: January 30, 2018

We are seeking a graduate RA to work on applying machine learning to understand characteristics of errors in GPU-based programs. The project, funded by the Department of Energy, involves understanding and then mitigating how errors propagate in GPU software, due to resource constraints on the GPUs and the use of mixed precision arithmetic.

Qualifications: PhD or MS (thesis) student in ECE or CS

Necessary knowledge/skills: System building skills (Python, Java, C), Basic exposure to data analytic and ML techniques, Computer architecture.

Desired knowledge/skills: CUDA-based GPU programming

The interested student is asked to submit an inquiry over email to Prof. Saurabh Bagchi and Dr. Paul Wood with the following items of information. Use as email subject “DOE GPU RA application”.

1) Department, MS/PhD, Year of graduate study

2) Proficiency level in the knowledge/skills mentioned above (high/medium/low)

3) Any prior research experience - mention publications (if any), time period, and supervisor

4) Grades in programming/algorithms/systems classes at Purdue along with when taken

5) CV in pdf format

Interviewing for the position will start beginning February and we expect to finalize the hire within 2 weeks.

Contact Emails: sbagchi@purdue.edu; pwood@purdue.edu