Podcast Summary: Lead With AI

Episode: Former Docker VP Reveals the AI Security Gap Costing Teams 80 Hours
Host: Dr. Tamara Nall
Guest: Jesse Williams, Co-Founder and COO, Jozu
Air Date: December 9, 2025

Episode Overview

This episode of Lead With AI features Jesse Williams, former Docker VP and now the COO and Co-Founder of Jozu, an AI DevOps platform. Dr. Tamara Nall digs into how Jozu addresses critical gaps in AI security and governance that leave many enterprises vulnerable—and wasteful. They discuss the importance of robust standards in machine learning (ML) deployment, real-life business impacts, and the future of operationalizing AI, focusing on practical solutions over hype.

Key Discussion Points & Insights

1. Jesse’s Background and Motivation

• Tech Experience: 15 years in tech, roles at Red Hat, IBM, AWS, VP at Docker.
• Personal Drive: “At my core, I’m a dad. I love being a dad. … My motivation is all about my kids, just providing for them.” (Jesse Williams, 02:06)
• Finds excitement and ongoing engagement in leading-edge tech.

2. What is Jozu?

• Core Focus: Helps enterprises get ML models into production quickly and securely, aligning with existing software standards.
• Problem Addressed: Traditional software governance frameworks didn't keep pace with the rapid adoption of AI/ML, creating risky security gaps.
• Innovation: Developed KitOps, an open-source project for artifact-based AI governance, now used by Red Hat, PayPal, ByteDance, and more.
  • “All the things that we’ve done to make sure our applications are trustworthy are non existent, even down to some of the most basic things like application versioning.” (Jesse Williams, 04:11)
  • KitOps is now under CNCF and has inspired the Model Pack spec.

3. The "Holy Smokes" Moment for Users

• Key Realization: Most teams think they have versioning handled, but Jozu reveals a single, tamper-proof artifact is missing.
  • “When we show them that open source project, Kit Ops, and we say, hey, this is all going to go into one artifact that is tamper proof… That’s where the light bulbs start going off and people start thinking, wow, I can do ML just as easily as I can do a microservices application.” (Jesse Williams, 06:16)

4. Under the Hood: How Does Jozu Work?

• Central Registry: Functions like ECR or Docker Hub but tailored for ML.
• Unique Add-ons: Handles diffs for prompts (critical given their length and complexity), package inference, effortless model deployment via Docker, and Kubernetes compatibility.
• Security Layers:
  • Integrated open-source security tools into a unified interface.
  • Cryptographically signed, tamper-proof audit logs—huge time savings.
    • “We’ve heard teams that take two developers two to three weeks to create a full audit log of their model and Jozu services that automatically.” (Jesse Williams, 09:54)

5. Practical, Even for Non-Developers

• Ease of Use: Allows non-technical users to pull models from Hugging Face and package them with minimal commands.
  • “For me, as someone who’s not super technical, I wouldn’t have been able to probably do that without significant help from ChatGPT.” (Jesse Williams, 11:44)
• Tool-Agnostic: Works with many model types, not just LLMs.

6. Naming Story

• Jozu: Japanese for “skilled or tactful,” inspired by co-founder Brad’s time in Japan. (13:04)

7. Ethics & Governance in AI

• Focus is on robust process controls, not just outputs.
• Key example: Preventing sensitive data (e.g., credit cards, medical info) from being included in models; compliance with HIPAA, GDPR.
• “We put governance into place so that … the things composing this model [are] the right things or not.” (Jesse Williams, 14:42)
• Considers cross-company and international data boundaries, model licensing, and embargoed regions.

8. The Future of AI Operations (MLOps)

• Emerging Trend: Shift toward on-premises and open-source AI models.
• Vision: Internal teams operationalizing critical AI workloads with Jozu, not just surface use-cases like chatbots.
  • “Our ideal future … is actually teams starting to operationalize ML inside of their business… not just having a support chat bot but internal, really important, really secure data.” (Jesse Williams, 17:28)
• On-Prem vs. Cloud: On-prem remains crucial for high-security, internal business. The main drawback is more in-house maintenance.

9. Getting Started and Community

• Try KitOps: Open-source, vibrant Discord community.
• Jozu Demo: Free sandbox at [jozu.com], with full feature POCs available.

10. Personal Insights & Bonus Segment

From One Genius to Another (19:56)

• Cherished Memory: Learning to ride a bike alone as a child. “I felt like I was flying.” (Jesse Williams, 20:09)

Rapid Fire Round

• Most Overrated AI Trend: “996 culture” (working 9am-9pm, 6 days/week) – “You get most of your creativity outside of the workplace.” (Jesse Williams, 21:22)
• Most Underhyped AI Trend: Custom GPTs—deeply training a GPT for a specific skill or task.
• Book Recommendation: Hackers & Painters by Paul Graham. “It’s also just a beautiful book, the way that it’s written.” (Jesse Williams, 22:43)
• Boldest AI Prediction: AI will enable more personal entrepreneurship. Predicts rise of billion-dollar 10-person companies.
  • “AI is going to open up the opportunity for everyone to basically be their own entrepreneur if they want to.” (Jesse Williams, 23:20)

Notable Quotes & Moments

• On security gaps in AI deployment:
  “ML came around so quickly that a lot of those things that we depended on to make sure our company doesn’t end up in the news for doing something bad … kind of got skipped and all of a sudden AI was here.” (Jesse Williams, 03:46)

• On why open source matters:
  “We built [KitOps] on open source … now it’s been adopted by companies like Red Hat, by PayPal, by ByteDance, the TikTok creators.” (Jesse Williams, 04:41)

• On personal creativity:
  “You get most of your creativity outside of the workplace, so work hard nine to five, put 40 hours in and you’ll be good.” (Jesse Williams, 21:30)

• On AI’s impact on opportunity:
  “I think it’s actually going to open up more opportunities for people.” (Jesse Williams, 23:37)

Important Timestamps

• 02:06 — Jesse discusses career, family, and motivation
• 03:05 — What Jozu does and why it was started
• 04:11 — The missing standards in ML security and governance
• 06:16 — Describing the “aha!” moment for users
• 09:54 — How Jozu automates audit logs, saving weeks of developer time
• 11:44 — Ease of packaging models without deep ML expertise
• 13:04 — The origin of the name Jozu
• 14:42 — Jozu’s approach to ethics and data governance in AI
• 17:28 — Vision for on-prem and future AI workloads
• 19:56 — Cherished childhood memory: learning to ride a bike
• 21:07-23:39 — Rapid fire (overrated/underrated trends, book rec, bold prediction)

How to Experience Jozu

• KitOps Project: kitops.org – free, open source
• Jozu Platform & Sandbox: jozu.com – demo access
• Connect with Jesse: LinkedIn (Jesse Williams), email (jesse@jozu.com)
• Community: Vibrant Discord for devs and users

Tone and Takeaway

Warm, down-to-earth, and practical—this episode avoids hype, focusing on authentic stories and significant, real-world challenges in AI governance and deployment. By the end, even non-technical listeners understand the urgency and impact of AI security gaps, the practical solutions Jozu provides, and feel inspired by Jesse’s vision of an AI-powered, opportunity-rich future.

For more conversation, insights, and firsthand innovation stories, subscribe to Lead With AI and join the discussion shaping the future of artificial intelligence.