Suggestions
Zohar Babin
Dreaming, breaking and making - Chief Business & Innovation Officer at Kaltura (KLTR: NASDAQ)
Zohar Babin is a veteran technology executive, platform strategist, and open innovation advocate known for building products, companies, and communities at the edge of what’s next. He currently serves as Chief Business & Innovation Officer at Kaltura (NASDAQ: KLTR), where he leads the company’s "Special Ops", focused on strategic growth, M&A, platform innovation, and launching entirely new business verticals. He is known as the Chief Troublemaker, a title earned by consistently building what’s next, moving fast, and reinventing the status quo.
Zohar has been with Kaltura since its earliest days, helping scale it from a scrappy open-source disruptor into a billion-dollar global video platform used by Fortune 100 enterprises, Ivy League universities, and major media and telecom companies. He has led the evolution of the company’s developer ecosystem, architected and launched its Video Platform as a Service (VPaaS), and navigated global security and privacy regulations while heading Kaltura’s platform, cybersecurity, and compliance operations.
His leadership blends technical depth, creative agility, and strategic investment foresight. Over the years, he has: Architected and scaled Kaltura’s OEM and platform business Built and empowered its global developer and open-source ecosystem Led security, tooling, and policy development for global customer compliance Served as Data Protection Officer, reporting directly to the Board and Audit Committee Launched new product lines in healthcare and executed multiple strategic partnerships and acquisitions
A Self-Taught Technologist Zohar’s career defies convention. He followed a hands-on, experience-driven path shaped by early entrepreneurship, military service, and self-learning. By his teens, he was running a tech consulting firm, online communities and tech startups, building software, training users, and designing real-world systems. He served in the IDF as a tank commander in the Armored Corps followed by a special strategic planning & intelligence role, where he led projects involving tactical 3D simulation and mapping, operational planning, and cross-unit coordination.
Builder and Innovator A lifelong builder and programmer, Zohar is a prolific open-source contributor in many projects. His work explores the convergence of AI, media, and automation, including: AI-infused video editing tools Enterprise-grade agentic copilots Conversational interfaces for personal and organizational memory systems He codes, prototypes, and architects systems designed for real-world utility and long-term scalability.
Industry Presence and Advocacy
Zohar is a regular speaker at global conferences including FOSDEM, DeveloperWeek, AWS re
Key Themes Zohar Brings to the Table The power of open systems in business-focused and proprietary domains Designing with ethics, agility, and community in mind Building with intent and grit Turning vision into product and code into revenue Why being a troublemaker is exactly what modern enterprises need
Zohar lives in New York City with his wife and two boys. Offline, he is either dreaming up the next tool that should exist, mentoring emerging founders, or proving to his kids that you can break the rules as long as you build something better in their place. He loves traveling, meeting people, and eating fantastic food. Chief Troublemaker. A dad who’s also a kid. Coding machines at night, programming people by day.
Highlights
Impressions from Apple WWDC Announcements Part 2: "OnDevice Processing" vs. "Private Cloud Compute"
𝐌𝐨𝐝𝐞𝐥𝐬:
- On-Device: ~3 billion parameters / 3.5 bit quantized / 49K vocab size. Designed to handle personal, low-latency, simple tasks efficiently.
- Private-Cloud: Undisclosed size, but significantly larger than On-Device / 100K vocab size. Designed to handle the more complex, broader context tasks. Apple claims strict access controls and privacy guarantees ensuring even Apple cannot access user data processed in the cloud (Apple states they welcome independent verification).
- Both types are designed to work on Apple silicon with lots of hardware-software integrated optimizations, Secure Enclave and end-to-end encryption (industry standard today).
- Both are designed with ephemeral data processing ensuring that data is not stored beyond the task processing duration.
𝐅𝐞𝐚𝐭𝐮𝐫𝐞𝐬 𝐫𝐮𝐧𝐧𝐢𝐧𝐠 𝐎𝐧-𝐃𝐞𝐯𝐢𝐜𝐞 𝐯𝐬. 𝐏𝐫𝐢𝐯𝐚𝐭𝐞 𝐂𝐥𝐨𝐮𝐝:
- Basic text tasks, contextual understanding of messages or emails. Probably small context window, so no large documents or video summarization tasks.
- Smaller vocab compared to the Private Cloud implies the On-Device models will be focused at single language.. although 49K is not that small (GPT3 had ~50K).
- Creating Genmoji, basic photo retouching / enhancement tasks - possibly On-Device, with more advanced Image processing and generation being "bursted" into cloud.
- "Siri enhanced" - simple conversational and local search including simple cross-apps tasks, probably On-Device, whereas more complex larger-context tasks (and external search based tasks) are likely to be cloud based. Onscreen awareness is a big one considering Windows Recall concerns, it will be interesting to see implementation architectural details.
- Voice memos transcription will likely be On-Device, with more complex cross-memos summarizations and insights requiring cloud.
- Accessibility features (Eye Tracking, Vocal Shortcuts) - likely On-Device only.
- ChatGPT Integration - on OpenAI's cloud.

Or why we shouldn’t necessarily design AI based on human inspired architectures ;-)
