2024 →
Founder
Cabinet
Open-source platform for running strategy execution on files you own.
Tel Aviv, Israel
Founder
Hila Shmuel
Founder, Cabinet. Previously Engineering Manager, AI & Data Infrastructure at Apple.
I spent four years at Apple building the infrastructure behind on-device AI: the training and data systems that produce the models running on billions of iPhones, Macs, and iPads. Twenty years of engineering led me to one conclusion: systems that scale to billions are the ones that hide complexity, not display it.
Cabinet is the application of that idea to a different problem: running strategy execution at the scale of an enterprise, on files the enterprise actually owns.
20+
Years engineering
4 yrs
At Apple, AI & Data
2
BSc degrees
The work at Apple
Four years building the infrastructure behind on-device AI.
I led the team that built the orchestration platform behind Apple's on-device AI training. Machine-learning engineers wrote Python. The platform handled fault tolerance, scheduling across thousands of GPUs, and the network topology across data centers. Training and data preparation became more than ten times faster, at a scale where ten times faster meant tens of millions of dollars and many engineering quarters returned.
Two consumer features shipped on pipelines I built end-to-end: Night Mode Portrait and Cinematic Video on iPhone 13. Both required new data infrastructure between research and shipping: collaboration across teams, careful trade-offs between quality and latency, and ML that had to feel reliable in a hand holding a phone.
What I learned at that scale is what informs Cabinet now: the interface to a powerful system should be a few sentences. The complexity behind the interface should be invisible. The work of the system should be observable and reversible.
Why Cabinet
Why I'm spending the next decade on this.
Every CEO has a Chief of Staff. Every SVP has an executive coach. The remaining 98% of managers have neither. That is not a tooling problem. It is an infrastructure problem. The work a Chief of Staff does (scan the room, surface what slipped, draft the reply, queue the meeting) is the same kind of invisible, scheduled, distributed work I built infrastructure for at Apple.
Cabinet is built the way good infrastructure is built. The executive writes a few sentences. The platform schedules, retrieves, drafts, and surfaces. The output lands in Slack, Teams, and email, wherever the executive already lives.
And the data (your strategy, your decisions, your one-on-one history) stays as plain files inside your company. Because the second lesson of Apple-scale was simpler: when you are responsible for billions of users' data, you do not trust anyone else to hold it. You hold it yourself.
The path
Twenty years going down to the metal, and back up.
2020 – 2024
Engineering Manager · Software Engineer
Apple, AI & Data Infrastructure
Led orchestration for distributed training and on-device ML data pipelines. Shipped Night Mode Portrait and Cinematic Video.
2022 – 2023
One-year career break
Road trip: North America, then Asia
Drove the United States and Canada in a van, then continued through Japan, Korea, Taiwan, Thailand, Vietnam, India, and Singapore. Returned with the perspective that launched Cabinet.
2017 – 2019
Senior Researcher
Armis Security
Vulnerability research and reverse engineering of embedded network devices. Built a network-crawler and configuration system from layer 2 up.
2011 – 2014
Software Engineer & Researcher
Israeli Defense Forces, Military Intelligence
Embedded systems, reverse engineering, C/C++ on Linux and Windows. Owned projects from idea to deployment.
Israel's first Hebrew podcast on bootstrapped and AI-powered startups.
Every episode is a conversation with a founder who built a revenue-generating company without venture capital. The infrastructure choices, the actual numbers, the pivots that saved the business. It is also how I keep my finger on where AI workflows break in production, and where the next leverage point is hiding.
Two BSc degrees, deliberately: Computer Science and Mathematics & Physics.
I finished my Computer Science BSc during high school, then read Mathematics and Physics afterwards. The depth and the breadth, taken in the order that gave me the most use of each. I have been writing software professionally since 2006.
Get in touch