Computer Architect - Massively Parallel Dataflow

Sorry, this job was removed at 6:35 a.m. (CST) on Thursday, March 17, 2022
Find out who's hiring in Austin.
See all Developer + Engineer jobs in Austin
Apply
By clicking Apply Now you agree to share your profile information with the hiring company.

About us:

Mythic has developed a unified hardware and software platform featuring its unique Mythic Analog Compute Engine (Mythic ACE™) to deliver revolutionary power, cost, and performance that shatters digital barriers preventing AI innovation at the edge. The Mythic Analog Matrix Processor (Mythic AMP™) makes it much easier and more affordable to deploy powerful AI solutions, from the data center to the edge device. The company has raised over $165.2 million to date, has over 120 employees, and has offices in Redwood City, CA, and Austin, TX.


About the role:

Mythic’s chip architecture uses a massively parallel dataflow strategy to keep Mythic’s special sauce, the Analog Compute Engines, at full capacity at all times. This requires efficiently tracking thousands of data dependencies in order to schedule hundreds of threads across dozens of tiles. To promote maximum performance and minimal data movement, all control is local or point-to-point, which provides unique synchronization and data integrity challenges. Mythic’s system combines concepts from out-of-order processing and distributed memory supercomputing to meet these challenges. 


The Computer Architecture team:

- Proposes architectural innovations to improve performance and energy efficiency on our next-generation products on those target networks.

- Through modeling, determines the opportunities for power and performance improvements for current networks, as well as the needs and challenges of future applications.

- Collaborates with the design and compiler teams, to define the appropriate chip-level and block-level specifications for ease of compiling networks and ease of design and verification, while still attaining our cost, performance, and energy efficiency goals.

- Reviews the implementation specification and verification and validation plans to assure that they match the intent of the functional specifications.

- Supports the execution from design start through production on architectural and use case-related questions.


This role is not location-specific. However, candidates will be expected to keep working hours in line with either Pacific or Central Time.

Here is what you will do:

  • Investigate and model distributed dataflow strategies in order to eliminate execution bubbles, bottlenecks, and deadlock. In particular, how to efficiently track thousands of dependencies and schedule hundreds of threads across a many-tile system.
  • Look for opportunities to streamline dataflow execution and minimize energy consumption.
  • Analyze the requirements of state-of-the-art neural networks and how they map to distributed dataflow systems.
  • Work with other technical leaders at Mythic to build a strong chip architecture
  • Engage with the Product & Business Development organization to prioritize new features and target applications.
  • Communicate a clear view of schedules, risks, and opportunities to company leadership
  • Stay up-to-date with the state-of-the-art for neural networks acceleration

Here is the background we hope you have:

  • Bachelors / Masters / PhD degree in Computer Science or Electrical Engineering
  • PhD + 5 years of industry experience or BS/MS + 10 years of industry experience required in chip architecture and modeling. This experience can include time spent in academic work, including papers, thesis, and dissertation work. The years required can be reduced if the work was related to systems for AI.
  • Academic and work knowledge of CPU, memory system, and bus interconnect architecture
  • Proficiency and experience with C++ / SystemC, as used in modeling for architectural performance
  • Proficiency and experience with spreadsheet and scripting based estimation of architectural cost, performance, and energy efficiency
  • Strong communication skills, both written and spoken

The following would be nice to have, but not required:

  • Published research or industry experience with out-of-order or distributed memory systems.
  • Experience working at startups
  • Experience with neural network frameworks (PyTorch, Tensorflow, ONNX)
  • Knowledge of the state-of-the-art in neural network architecture



At Mythic, we pride ourselves in creating a culture where all employees feel valued and appreciated for the diverse perspectives and backgrounds they bring to the team. We aim to hire smart people, give them the resources they need to do their job well, and then leave the rest up to them. We celebrate individual differences and encourage people to be comfortable bringing their authentic selves to work. At the end of the day, we are committed to building a diverse workforce where everyone belongs.


Mythic is an equal opportunity and affirmative action employer. It ensures equal employment opportunity without discrimination or harassment based on race, color, religion, sex (including pregnancy, childbirth, or related medical conditions), sexual orientation, gender identity or expression, age, disability, national origin, marital or domestic/civil partnership status, genetic information, citizenship status, veteran status, or any other characteristic protected by law.


We look forward to reviewing your application!

Read Full Job Description
Apply Now
By clicking Apply Now you agree to share your profile information with the hiring company.

Location

Our office is located close to The Domain, where there are a variety of shops and dining options.

Similar Jobs

Apply Now
By clicking Apply Now you agree to share your profile information with the hiring company.
Learn more about MythicFind similar jobs