Ph.D. Position in in-memory computing

Posted on

Catégories:

Open positions

Your mission:
In-memory computing is a particularly interesting area for mixed-signal circuit designers as it is where analog meets digital almost in a literal sense.  
 
Many and many in-memory computing demonstrations have come out in the past several years – at a record pace reflecting on my past 15+ years of observation. We all now understand that the accuracy of in-memory computing done in the analog domain would never pare to 32-bit floating point digital implementation, but there are interesting edge applications where such inaccuracy is very well tolerated. Also, the efficiency of those implementations is quite astonishingly high as it breaks the so-called ‘memory energy wall’. 
 
Now, thinking into the imaginable future where all edge-facing memory products would incorporate the in-memory computing feature into them while it introduces the least disruption of the memory function itself, we must think about what kind of in-memory computing architecture is the least invasive, mass-production friendlym and also has the least overhead for the memory density. For this, we look nowhere further than what we do excellent in the MSIC Lab – a charge-domain circuits for in-memory computing.
Main duties and responsibilities include :
Develop an array-scalable and robust charge-domain architecture for in-memory computing circuit
  • Study the charge-domain ADC design paradigm and suggest a novel architecture optimized to the in-memory computing application
  • Design a custom SRAM array at the transistor level
  • Design peripheral circuits for the in-memory computing
  • Architect the data flow for the in-memory computing accelerator
  • Apply an inference model to run on the in-memory computing hardware
Your profile:
  • Solid background and/or experience in mixed-signal circuit design (this means switched capacitor circuits and not RF)
  • Excellent communication/presentation skills in English
  • Excellent knowledge of the following topics:
    • Charge-domain circuits
    • Dynamic amplifiers
    • Analog to digital converters
    • Energy efficiency
    • integrated circuits design flow:
      analog design, digital synthesis, layout, place and route…
  • Necessary skills:
    • Fluency in using Cadence Virtuoso Suite
    • Fluency in implementing inference models in Python
    • Fluency in using SPICE simulators (HSPICE, FineSim, AFS, etc.)
    • Fluency in using PCB design suite (Altium)
  • Preferred skills
    • Fluency in running simulations on the Linux terminal 
We offer :
  • A multidisciplinary environment
  • Excellent working conditions
  • Laboratory based in Neuchâtel campus (our building is quite new!)
Preferred start date :
January 1, 2023 +/- 2 months
 
Duration :
For Ph.D: 1.5 year, will continue with a different project
 
Contact :
Candidates must send their application to Prof. Kyojin Choo