Project Details:

<aside> <img src="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/fc3ec986-048e-411a-a509-a3e164f7ab2f/31493405.jpg" alt="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/fc3ec986-048e-411a-a509-a3e164f7ab2f/31493405.jpg" width="40px" /> Preferred field of study: Electrical Engineering / Computer Science

Remote work: not preferred

Project Size: 4

Expected salary: 27,50 EUR/h

</aside>

About Recogni

<aside> <img src="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/38531361-b605-4845-aa67-eca5471d6626/31493405.jpg" alt="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/38531361-b605-4845-aa67-eca5471d6626/31493405.jpg" width="40px" /> About Us

Recogni was founded in 2017. It’s split 50-50 across San Jose, CA, and Munich, Germany. We are developing AI silicon – with a current focus on autonomous vehicles (autonomous driving, assisted driving functions etc.). Chip, hardware and firmware design are done in San Jose, while all AI-related, compiler and devops engineering happens in Munich. Our engineering runs on the philosophy of HW-SW-Co-Design.

</aside>

Your Tasks & Project

<aside> <img src="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/cf5ea018-da42-4f46-a9e3-b3d7a9086ad2/31493405.jpg" alt="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/cf5ea018-da42-4f46-a9e3-b3d7a9086ad2/31493405.jpg" width="40px" /> Expand current evaluation landscape:

AI chips (such as the ones by Recogni) are designed to efficiently and numerically accurately accelerate various ML workloads. To get the most out of an application (e.g., “object detection in an autonomous vehicle”), there is a multi-dimensional optimization landscape that an ML engineer has to consider for deploying a model on an edge device: Model architectures–some architectures are a better fit to Recogni’s AI chips than others. Model quantization/compression methodologies–how to most efficiently and effectively convert ML models to a specific AI chip’s math is an active field of research. Chip design parameters–our current but especially also our next-generation chip (which is currently under development) have a number of knobs to tweak and tune which can have a huge influence on how suitable an architecture is. We expect the group of students to expand Recogni’s current evaluation landscape in all three dimensions. This means: Working closely with our Perception team to get more and more off-the-shelf state-of-the-art machine learning models integrated into our ecosystem so that we can run evaluations on them. Working closely with the Core-Tech team to understand how networks are being converted today, and what further new methods might be worth being integrated into the evaluation framework to expand the search space and potentially find better ways to teach a network how to operate accurately with our chip’s math. Working closely with the Core-Tech and Silicon Design teams to enable a wide search over many different number system parameters (for example: instead of having a 14-bit accumulator for our convolution we might want to test 13-, 15-, and 16-bit accumulators).

</aside>

Your Benefits

<aside> <img src="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/d84abd94-2e68-415c-ad01-f98388e6807b/31493405.jpg" alt="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/d84abd94-2e68-415c-ad01-f98388e6807b/31493405.jpg" width="40px" /> Compensation & Benefits

Work within a multinational, strongly-funded deep-tech startup with roots in both Silicon Valley as well as Munich
Work among word class talent in AI and chip design
Above market average compensation
Beautiful new office space close to the central station (former Google office) with all kinds of nutritional benefits for free
Perspective to join Recogni’s engineering organization as an FTE after graduation

Expected salary: 27.50 EUR/h

</aside>

Your Superpower

<aside> <img src="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/2221a94d-9943-48cf-8e70-338dfb9b8d3c/31493405.jpg" alt="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/2221a94d-9943-48cf-8e70-338dfb9b8d3c/31493405.jpg" width="40px" /> Required Skills & Qualifications

EE (or CS, but preferably the first)
Experience in a major ML framework (PyTorch or Tensorflow), and designing/training neural networks
Good intuition for low-level digital design choices (bits and bytes) and their influence on high-level performance
Experience in structured SW engineering and reporting </aside>

Further Information

<aside> <img src="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/8c4c5b06-712b-45d3-af4e-f9d04f077b23/31493405.jpg" alt="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/8c4c5b06-712b-45d3-af4e-f9d04f077b23/31493405.jpg" width="40px" /> Additional Questions

Am I allowed to work remotely and abroad? Not preferred. We have a big, beautiful office with lots of space and would ask the students to ideally come in two full days a week.
Am I allowed to have a second job in parallel? (e.g. Minijob, Tutor / Research Assistant at university, etc.)

Is allowed, you only have to inform us about your further employment and you have to make sure that you don't exceed a weekly working time of 20h/week due to social insurance regulations.

</aside>