Roei Kisous

Software Engineer @ Google Cloud

🌐 roeikisous.com

M.Sc. in Computer Science from the Technion with vast experience from robust backend systems in cloud security to high-performance storage infrastructure.

View Concise Resume

About

I am a Software Engineer currently working in Cloud Security (SecOps) at Google, focusing on building scalable, reliable services and leveraging AI to enhance workflows.

Alongside my work at Google, I collaborate with the Technion as a freelance researcher. Through this ongoing research and my previous industry roles, my work spans data migration algorithms, near-memory processing, optimizing storage for AI infrastructure and SSD garbage collection algorithms.

I enjoy tackling complex, system-level challenges and building practical, high-performance solutions.

Experience

2025 — Present

Software Engineer, Cloud Security (SecOps)

Google · Tel Aviv, Israel
  • Leading the end-to-end integration of a GCP-based case management service for SOAR with a new internal Google service built on Borg. This cross-organizational initiative (VP Announcement) significantly enhances system scalability, reliability and functionality.
    • Tech-lead of the high-throughput synchronization flow utilizing GCP Pub/Sub to bridge cloud-based case management with internal Borg services, ensuring high-fidelity data consistency globally.
    • Driving the development of a streamlined case management experience (handling cases, alerts, and investigations) to unify security operations (Release Notes).
  • Part of the Threat Hunt Agent initiative, leveraging AI models to automate and facilitate advanced threat hunting capabilities (featured at Google Cloud NEXT 2026: Video · Blog).
2025 — Present

Freelance Researcher

Technion – Israel Institute of Technology
2022 — 2025

Software Engineer, Cloud Storage

Toga Networks (Huawei) · Tel Aviv Research Center
  • Conducted next-generation research for AI infrastructure acceleration, specializing in storage performance optimization. Received the Huawei Future Star Award (2024).
  • Contributed to a high-performance key-value embedding store for AI, optimizing KV-Cache for large-scale machine learning models.
  • Implemented a new architecture for Huawei Cloud's file system by transitioning to FUSE over virtiofs and offloading the backend to DPUs. This architectural shift reduced resource usage by 75%, increased throughput by 14x, and improved latency by 50x (Poster).
  • Designed a near-memory processing architecture for block devices, achieving up to 25x speedup over conventional remote memory architectures (Poster).
2019 — 2020

Software Developer

Deep Learning Robotics · Caesarea, Israel
  • Developed a vision-based robotic controller using C++ and OpenCV, enabling robots to learn and replicate complex tasks through human observation.

Education

2020 — 2022

M.Sc. in Computer Science (Storage Systems)

Technion – Israel Institute of Technology

GPA: 91.4

Thesis: Clustering Based Data Migration in Deduplicated Storage (Advised by Associate Professor Gala Yadgar).

Publications: FAST '22, ACM TOS '22, SYSTOR '22 Highlight

Served as teaching assistant in charge of the Database Systems course, overseeing all aspects of the course including recitations, homework assignments and exams, for ~200 students per semester.

2018 — 2020

B.Sc. in Computer Science

Technion – Israel Institute of Technology

GPA: 85.3

Multiple Dean's List awards.

Publications & Patents

MASCOTS 2026 Workshop

MetaGC: Improving SSD Garbage Collection with Randomization

A randomized meta-algorithm for SSD garbage collection that dynamically filters high-quality candidate blocks. This approach improves victim selection tradeoffs and seamlessly integrates with existing policies to drastically reduce scanning overhead without impacting write amplification.

Read More
Patent (Pending)

Hardware AI Patent

Currently pending. Focuses on hardware-accelerated AI infrastructure and novel storage architectures.

Read More
SYSTOR '23

Near-Memory Processing Offload to Remote (Persistent) Memory

Enhances remote memory architecture using Near Memory Processing (NMP) by offloading specific compute tasks from the client to the server side. This approach significantly saves network bandwidth, reduces round-trip times, and minimizes distributed lock overheads.

Read More
ACM TOS '22

The what, The from, and The to: The Migration Games in Deduplicated Systems

Expanded journal version of the FAST '22 research. Details three distinct algorithmic approaches: greedy, theoretically optimal ILP, and clustering, to optimize data migration in deduplicated storage while balancing computation time, efficiency and network utilization.

Read More
FAST '22

The what, The from, and The to: The Migration Games in Deduplicated Systems

Formulates the general data migration problem for deduplicated systems as an optimization challenge. The proposed algorithms, greedy, theoretically optimal ILP, and clustering, minimize system size while ensuring even storage load distribution and bounding network traffic.
Best paper runner-up · SYSTOR '22 Highlight

Read More