Associate Professor TU/e (JADS)-NXP - Predict HPC work-loads for EDA projects

Engineering Data and AI Supremacy for the Semiconductor Industry: A Research Agenda

In a world that is getting smarter (smart cities, smart cars, etc.) there is a growing demand
for chips (semiconductors) that provide products with intelligence and connectivity, in a
secure way. Several markets (automotive, consumer electronics) are severely impacted by
shortage of chips. NXP is an R&D company that exploits latest technologies (16 nm FF, 7 or
even 5 nm) with their innovations. Time criticality is absolutely key for being successful in
this business, and total quality. That’s why there is a growing need for high performance
compute capacity to run Electronic Design Applications (EDA).

Interestingly, it is expected that there will be nearly 10x the number of connected devices in
existence by 2030. This will pose a huge challenge to companies like NXP. Those connected
devices and applications include for example: smart home/work/city, 5G, Wearables, Self driving
cars, and robotics.

Broadly speaking and according to McKinsey, semiconductor companies could benefit by
rethinking their approach in six critical areas: technology leadership, long-term R&D, resilience,
talent, ecosystem capabilities, and greater capacity. At the same time, with greater competition,
semiconductor companies would be wise to increase their efforts to recruit talent, including staff
with expertise in process technology and operations management.

Broad-Spectrum Collaboration Opportunities

Quoting from the same article, three foci for joint research and exploitation emerge:

• Ecosystems’ standardization + sustainability analysis, synthesis, and automated support. “In another type of collaboration, companies may create ecosystems in which one player develops intellectual-property (IP) blocks that many customers can leverage. Arm, for instance, has developed an architecture for a processor that others may license. This strategy decreases costs for all involved. Some companies have also formed strong IP partnerships with academic institutions.”

• Service value-chain analysis and automated sustainability support. “Major semiconductor players also have a long history of joining forces to develop and align their technology blocks, thereby reducing the chance that one organization will create a technology that does not fit into the value chain. Similarly, industry associations can play an important role in providing guidance on the long-term technology road maps, and dedicated semiconductor research institutes, such as Imec, in Belgium, may convene players to collaborate during precompetitive research.”

• Strategic innovation services and data products. “Beyond partnerships, semiconductor companies might want to undertake a programmatic M&A strategy—a serial approach to smaller acquisitions, along a specific theme—as the industry consolidates. If they do so, they might benefit by focusing on acquisitions that would allow them to branch into adjacent areas, open important markets, or add capabilities essential for future growth and for extending their technology leadership. The current scarcity of targets, however, requires potential acquirers to investigate and execute mergers quickly.”

Research Roadmap

The big data deluge is forcing a need for large-scale analytics applications which are bound to
be smart, namely (1) self-cognisant or able to automatically adapt to growing data
complexities, e.g., by integrating seamlessly and securely diverse computing and data
environments, spanning from core cloud to edge; (2) self-sustaining, or able to intelligently and
autonomously address changes in application infrastructure and data variability offering
automatic (re-)deployment, mobility, and secure adaptability of services from cloud to edge to
diverse users and contexts; (3) self-restraining, or able to—stemming from stringent needs to
address low resources-consumption and waste reduction—steer resource management while
taking into account the openness and (un-)trustworthiness of the underlying resource
management layers.

Such smart operations are therefore required to interface with all the layers in the computing
continuum plane and will likely be required to learn through the monitoring and management of
resources deployed on Cloud/Edge while at the same time applying AI-techniques that cater for
dynamic load balancing to optimise energy efficiency and maintaining balanced data traffic and
high, distributed, reliable throughput from cloud to edge according to the application and user
needs and the underlying infrastructures (e.g., considering varying supply of green energy to
optimise environmental footprint).

As an example of this challenge, a software system offering a functionality built on a
neural-network may only be effectively assessed in a NXP production environment and it is
therefore difficult to test in a traditional software development cycle, because engineers do not
dispose of a training sandbox that is representative of the scale and diversity of production
inputs as well as of the conditions reflected in the above challenges. This triggers a fundamental
challenge for software engineers, namely how can functional correctness and quality-of-service
be guaranteed in large-scale AI-enabled cognitive cloud analytics?

To realize the above vision, and get answers to similar questions as the question above, we
have defined the following key objectives:
● O1 - Engineer Friendly Development Platform : A low-code development platform to
efficiently assemble and deliver in production Large-scale AI-enabled analytics for EDA factored
into data-products;

● O2 - DevOps Data-product centric Methodology : An operational data-product centric
DevOps methodology, equipped with a governance model to foster higher quality-of-service and
sustainability-of-service in today’s software system development and operations. DevOps
should revolve around two production lines: one to design, develop and public data products,
and one assembly line to compose & test & provision them;
● O3 - Self-Adaptive AI-Enabled Analytics Platform : A machine learning enabled monitoring
toolchain and feedback actions to continuously drive the evolution of AI-enabled analytics
product based on empirical data on the application and the environment collected at runtime.

In order to maximize the impact of this research we wish to (a) work on concrete studies
delivered by NXP, (b) do the work in mixed teams of scientists (of all levels) and professionals, (
c) organize workshops and seminars to disseminate (intermediate) results, (d) compile a book at
Springer or MIT Press, (e) develop new joint EU level project proposals, (f) embark on joint
collaborations with other academic institutions like POLIMI (Milan).

To execute the above research agenda, we are looking for a part-time associate professor.
Tasks of Associate Professor:
1. Help to initiate, coordinate and supervise triple-helix involvement of MSc DSE students, e.g., through graduation projects, course assignments, guest lectures, etc.
2. Bootstrap Professional education: the joint associate will assist in specifically designing tailored professional education for NXP, e.g., he will serve as a liaison between the JADS professional education team and the NXP counterparts.
3. Setup, align and co-supervise/mentor EngD trainees. This includes setting up data challenges, and, guiding the 1-year final EngD projects.
4. Drive new R&D initiatives in high-risk/high-impact areas exploiting national and EU-level R&D programs.
5. Define, setup and co-supervise PhD trajectories as a series of short-cyclic, learning-by-doing based projects.

More information about NXP in the Netherlands...