About
Hi, I'm Marc.
I've spent thirty years in software and more than twenty running the engineering and product organizations behind it: cloud infrastructure at Rackspace, SoftLayer and IBM, digital banking at Alkami through its IPO, and AI making real decisions for insurers at Shift Technology. This year I started building products myself again, with AI agents doing most of the implementation. Marbir Digital is where that work ships.
Four platform shifts
I started out as a sysadmin and web developer in the dot-com boom, and I've been lucky to be near the front of every big shift since.
Cloud
I built infrastructure products through the move to cloud. At SoftLayer I founded and led the product innovation team and invented a patented method for deploying servers across virtual and bare-metal environments. After IBM acquired SoftLayer, I became CTO and Distinguished Engineer in IBM Cloud. Along the way I shipped several early App Store apps for managing cloud storage and CDNs from a phone.
Digital banking
Joined to lead product engineering, CTO from 2020. Took the company public in 2021. Moved more than 50 banks and credit unions onto AWS, built them a customer-facing data warehouse, and grew the platform from 5 million to 13 million users. All of it with strong security and compliance.
AI in production
At Shift, the product was the model. Our AI made fraud and claims decisions for insurers around the world, so when it was wrong, an insurer paid a claim it shouldn't have, or denied one it should have paid. As CTO in Paris I built the Insurance Data Network and put generative AI and LLMs into the product. Getting it right had far more to do with data, testing and governance than with algorithms.
Agents
At Shift and then as CTO at Pleo in Copenhagen, I made GitHub Copilot, Claude and Claude Code part of how 200+ person engineering organizations build. Then I went further and started building with agents myself, every day. It's the first shift that changed how I personally work, and I didn't think I could lead it well without doing that.
What I know from doing both
Twenty years of running the org, plus six months of shipping with agents, gives me a view most leaders only have half of. What I'd tell any CEO or board right now:
Review is the new capacity constraint.
In my own setup, Claude writes the code, a second model reviews every change, a third shadows without blocking, and nothing merges without my approval. Code arrives faster than anyone can read it. Plan senior review time the way you used to plan headcount, and tier it: AI takes style, tests and obvious bugs; people take intent, architecture and risk.
When building is cheap, saying no is the job.
To get Swarmers ready to ship, I cut about 80% of the planned scope. Engineering capacity used to do your prioritizing for you. Now someone has to hold the line, and that person needs to understand both what users want and what every addition costs to own. That's why I think product and technology alignment is more important than ever.
Agents made building cheap. Owning still costs the same.
For my first game I built a custom leaderboard backend with Apple and Google sign-in. For the second I used Game Center. Building the backend with agents was easy. Running it was the part I hadn't priced in.
Accountable AI starts with the data.
Shift taught me that. Clean data, real tests, clear ownership and an audit trail come first. Autonomy comes after, one category of decision at a time.
Agents are your newest integration customer.
Integrations were always the hard part. Now agents hit your APIs and MCP servers more often and more literally than any person. Testing a data vendor's MCP server for LORE turned up stale catalog data and a broken calendar endpoint. If you ship that layer, treat it as product.
What I'm building
Every Marbir Digital product is designed, built, shipped and run by me, with agents writing most of the code. My daily tools are Claude Code, Codex and Cursor.
Swarmers
A 1980s-inspired arcade defender for iOS, live on the App Store. The idea sat in a notebook from the first iPhone launch for almost twenty years. Learn more
Word Biome
A word game where every creature is a real animal, and tapping one tells you about it. Built to be family friendly, with a kids mode designed to help children learn as they play. Live on the App Store. Learn more
LORE Tennis
A tennis app that puts the players at the center. Most sports apps are organized around the sport: tournaments, draws, schedules and scores. LORE follows each player instead, with their ranking, stats, news, social posts and trading cards in one place. Built on NDX.
NDX
A backend data platform for trading cards. It serves the card catalog and pricing data underneath LORE, as a service other products can build on.
trvlr
A concept exploring how well AI can turn the travel posts you share into real places on a map. The foundation is data: AI models extract the places mentioned in videos, captions and posts, classify them, and match each one to a real point of interest using open data from Overture Maps and Foursquare, plus Google Places. See the concept
Work with me
Leading product and technology.
The seat where I add the most: CTO, CPO or CTPO at a growth-stage or PE-backed company where AI has to show up in the product and in how the team works.
Advising.
I work with CEOs, boards and PE sponsors on technology strategy, AI readiness, and what to ask of their technology leaders.
Find me on LinkedIn (Marc Jones and Marbir Digital) or at hello@marbirdigital.com.