Shayan Sadigh

Location:Santa Clara, CA
Email:ssadigh.contact@gmail.com
Github:https://github.com/shayanpersonal (200+ stars)

Summary

Founding Engineer, Agentic AI @Aurascape.ai | Previously @Amazon SageMaker
Placed top-10 in $1.5 million Kaggle competition sponsored by Department of Homeland Security. Designed and trained MVCNN neural net in PyTorch for threat recognition on 3D body scans. (Slides)
Authored "checkered subsampling" deep learning research paper, published PyTorch code. (arXiv)
Excellent teaching and presentation skills.

Founding Engineer, Agentic AI (Feb. 2024 - Current)

Addressing emerging security challenges. Built in-house agentic systems in active use by the team and product. Drafted patent applications and helped secure $12.8 million in seed funding.

Adjunct Faculty (Aug. 2023 - Dec. 2023)

Taught in-person 16-week course on Python, machine learning, and deep learning. Labs included a "small language model" neural network trained with PyTorch on tiny-shakespeare dataset. Incorporated MoeBot, my previously built agentic AI system, as an AI teaching assistant.
Quote from department chair: "The student evaluation ratings are the best I've ever seen out of the 12 years of evaluations packets I've signed off on..."

Entrepreneurship (2021-2023)

Built MoeBot, a subscription-based Discord AI agent powered by a swappable LLM backend. Incorporates RAG, agent memory, tool calling, Stripe payments. Deployed to Google Cloud.
Built various other projects such as MIDI recording tools and VST audio plugins.
Taught a college course and incorporated MoeBot. Pursued musicianship, explored my interests.

Deep Learning Engineer (Mar. 2020 - Dec. 2020)

Worked with researchers to improve AI inference in Amazon SageMaker's PyTorch fork.

Senior Staff Software Engineer (Oct. 2019 - Mar. 2020)
Staff Software Engineer (Sep. 2018 - Sep. 2019)

Awards: Project of the Quarter - Machine Learning Antivirus
Proposed and implemented new stack-based pattern matching algorithm in C which improved threat coverage in PAN-OS firewall, resulting in higher scores on 3rd-party benchmarks.
Solved many long-standing tickets from customers, data science, product managers, and threat research. Quote from Vice President of Engineering: "One-man army."
Made major contributions to industry-first machine learning features - integrated inline machine learning classifiers into resource-constrained firewall in C.
Prepared and gave presentation explaining the pattern matching system in the threat detection codebase. The presentation recording is now used to train new hires and cross-functional teams working with threat detection code.
Brought up and helped train two junior engineers.

stacked-autoencoder-pytorch (100+ stars) - a popular PyTorch autoencoder influenced by Hinton's 2006 "deep belief nets" paper.
Checkered CNN (30+ stars) - a deep learning technique I proposed, checkered subsampling.
Passenger Screening Model (20+ stars) - my top-10 solution to a $1.5 million Kaggle challenge.
GauchoMap Chrome Extension - 200+ users when I was maintaining it (2016).

5-year accelerated B.S. + M.S. program @ University of California, Santa Barbara
Master of Computer Science (2017 - 2018)
Bachelor of Computer Science (2013 - 2017)

2nd-degree black belt in Taekwondo. Practicing Brazilian Jiu-Jitsu and Muay Thai.
Playing and writing piano music since early childhood.
Expert Dance Dance Revolution player.
Ranked #1 in competitive online arena shooter.
Languages: English (native), Farsi (native, but very rusty)