Sergey Ostrikov
00xnor@gmail.com

Hi, I’m a systems engineer with a decade of hands-on experience in silicon development, including early-stage development of an AI inference accelerator and full-cycle work on a secure memory IC from concept to delivery and certification. I enjoy building prototypes that help stakeholders decide on new product commitments. Some things I’ve been up to professionally:

Lead Principal Engineer, Infineon Technologies, 2020-2025, San Jose, CA
In this role, I shaped the evaluation of an analog in-memory compute technology. I developed a graph compiler as a primary vehicle for architecture exploration and tradeoff analysis.
For PPA analysis, I developed a heuristic-based algorithm that maps a task graph onto a hardware connectivity graph (demo).
(neural network) (distributed in-memory compute fabric)
Formulated a nuanced approach for assessing accuracy of analog inference beyond standard evaluation metrics (description).
And finally, I did a comparison of the technology against leading AI accelerators and presented my findings at a JEDEC AI workshop.

Conference Talks
Graph Compiler for a Weight-Stationary Dataflow
JEDEC Mobile/Client/AI Computing Forum
🇺🇸 San Jose, CA, May 2024
In-memory Computing using SONOS (Non-Volatile Memory)
JEDEC Mobile/Client/AI Computing Forum
🇹🇼 Hsinchu, Taiwan and 🇰🇷 Seoul, S. Korea , May 2023

Resilient Boot (secure boot on processors without eNVM)
IEEE International Conference on Cyber Security and Resilience
💻 Virtual, July 2021

Systems Engineer, Cypress Semiconductor, 2014-2020, San Jose, CA
In this role, I led a feasibility study for a secure memory device, analyzed and improved secure memory IP, and contributed to successful security certification (CAVP, CMVP).
Wrote a C model of a secure NOR Flash device to determine security aspects of the product itself, along with the systems it was intended to operate within (available in the product SDK).
Coded apps, middleware and numerous drivers for demo purposes: MQTT, TLS, TCP/IP, WiFi, ETH, BLE, …

Publications
Resilient Boot
A practical guide for achieving FIPS-certifiable DICE attestation
Secure memories in a connected world (page 30)
How to design FIPS 140-2 cryptographic modules to meet TCG Implicit Identity Based Device Attestation

Patents
Memory device resilient to cyber-attacks and malfunction
Methods for updating firmware with single memory device
Secured communication from within non-volatile memory device
Access-aware flash translation layer on flash memory IC

Contributions to Standards
JEDEC: Serial NOR Security Hardware Abstraction Layer (throughout the doc)
TCG: DICE Attestation Architecture (chapter 7)
TCG: Cyber Resilient Module and Building Block Requirements (storage protection)