Use expand-and-contract to ship database migrations with confidence.
Challenge conventional wisdom before making hard decisions. Some one-way doors open both ways.
Synchronization barriers let you test for race conditions with confidence.
Neon Testing lets you test against real Postgres with zero infrastructure.
I founded STΛR MODΞ, a collective of independent builders. Everyone brings their own projects, from open source and developer tools to commercial products, and we meet on a weekly call to discuss each other’s work, share progress, and trade feedback. We publish what we’re building on starmode.dev.
Neon Testing is one of mine, a testing utility for seamless integration tests with Neon Postgres, for Vitest and Bun Test.
I’ve been building for AI teams for over a decade. Long before today’s LLMs, I built tooling for predictive analytics and machine learning. These days I architect and build agentic systems for venture-backed startups, VC firms, and enterprise clients.
Coupa is a leader in business spend management. They brought me in to set the architecture for their supplier agent and build it out. It’s an email-native AI agent on AWS Bedrock and LangChain, with evals guarding its behavior as it evolves. Suppliers email it the way they’d email a person, and it replies in the same email thread. It answers questions about invoices and purchase orders, and takes action too, like creating and sending invoices.
Lakebase is Databricks’ serverless Postgres. I’m designing their data anonymization workflow, where engineers branch production, mask the sensitive data with join-preserving transforms, and verify nothing leaks before restricted teams get access to the anonymized branch. It ships as an agent skill, so a coding agent can walk any engineer through it.
Neon is serverless Postgres. I build developer-facing AI tooling for them, including an open-source agent skill that helps coding agents catch costly egress patterns, evals that measure how agents perform with and without it, and reference examples pairing Neon branches with preview deployments.
Scoutbee builds AI-powered supplier discovery for procurement teams. I built Instant Supplier Search (ISS) as a skunkworks project, letting Fortune 50 buyers find and evaluate suppliers across any category with natural-language queries and rich filtering. I then worked on Aura, the next generation of the product. Scoutbee was later acquired by Coupa.
Mikael is a great developer with a natural interest and understanding in the business side of the applications, which makes it very easy to coop with him. He is fast, thorough and working always with bleeding edge tech. All I am looking for when working with a freelancer. Highly recommend Mikael!
BenAi is an AI-first chatbot for health benefits administrators (TPAs). They brought me in to take the platform from prototype to production. BenAi ingests plan documents and claims data at the individual member level, using LLMs and custom embeddings to give call center agents accurate, member-specific answers on claims and coverage, cutting call times and escalations.
I had the pleasure of working with Mikael on my company, BenAi, and he brought a whole new level of professionalism and polish to our software. From day one, he zeroed in on areas for improvement and took our platform from a prototype to a production-ready product of true enterprise quality.
What stood out the most was Mikael’s incredible attention to detail. He’s always up-to-date on the latest technology trends and consistently recommends modern, high-performing solutions that have significantly improved the functionality and performance of our platform. He didn’t just upgrade the code; he transformed our development process with a disciplined, organized approach that brought real structure to our team.
Mikael’s work has elevated BenAi in ways I hadn’t imagined. If you need someone who can turn ideas into robust, scalable solutions, I would highly recommend working with Mikael.
BootstrapLabs is a venture capital firm with an applied AI venture studio, where I led a team of four building Shorty, an AI-augmented communication platform. It organizes conversations and interactions into a graph, and relies on AI agents to identify consensus points and support decision making, bringing every conversation to a conclusion as quickly as possible.
It’s been a privilege to work alongside Mikael for the past 15 years. He’s not only an exceptional developer and mentor but also has an impressive talent for breaking down complex, emerging technologies like AI into insights that are accessible and easy to understand. His expertise and guidance have been invaluable, and we look forward to future collaborations with Mikael.
Teledyne FLIR is the world leader in thermal imaging cameras. Years before today’s LLMs, I built Conservator, their data lifecycle management platform for teams training neural networks. It takes video uploads in any format from FLIR offices around the world, splits them into frames, and provides annotation and versioned training datasets. Annotations made by the models come back into the tool for correction, closing the loop for the next round of training. What started as an internal tool is now sold as an OEM product.
Mikael was a pleasure to work with. Very knowledgable, very competent, very productive.
I occasionally speak at meetups, conferences, and private corporate events about leveraging AI to build better software. Want me to speak at yours? Get in touch.
AI dramatically accelerates software development, but how it is used often compromises quality and structure. Without a clear process, AI-generated code can become inconsistent and difficult to maintain. In my recent work, I’ve been developing a way to use AI to write high-quality, well-tested, and high-performing code really fast. This approach is inspired by my work building PimDB, a browser-based persisted in-memory database. In this talk, we’ll explore a structured process to empower teams to leverage AI to tackle complex projects with incredible velocity without compromising on quality.
We’ll cover: API design and architecture · Writing tests · Performance benchmarking · Writing documentation · Implementation and optimization
Attendees will gain practical strategies for using AI to write maintainable, reliable code at AI speed — without compromising on quality.
Mikael came to our company and gave a well-organized and highly informative talk to our developers about how to use AI to speed up software development in a practical way. Mikael has a knack for explaining complex ideas in a way that is very accessible, and our entire group was immediately inspired and empowered to use the emerging methodologies he showcased. I was super impressed!
My peers and clients often point out my attention to detail and excellent communication skills.
Mikael has been with us for more than a year and I can’t say enough good things about him! He’s an exceptional developer but what really sets him apart is his ability to think at the architecture and business level. Everyone at Evive has loved working with him and I consider myself lucky to have found him. If you get the chance to work with Mikael, take it. I look forward to working with Mikael again in the future.
We couldn’t have asked for anything more from Mikael. We had a relatively difficult new application which needed to be built, on a short timeframe and to specific requirements. His approach was careful, considered and fast throughout.
He is an excellent communicator and was able to work considerately, prioritise tasks appropriately and ask questions when needed. He understood the task well and he produced very high quality code, in line with all the best practices etc. - I am confident that his code will be running with us for a long time to come.
Thank you Mikael for a job well done, we’re looking forward to working with you again in the future.
Mikael is fantastic! We are already finding new ways to utilize his vast knowledge. HIGHLY RECOMMEND!
Mikael did a great job with our project. He provided excellent documentation and handed over all of the necessary information.
He also worked with my to make sure my AWS S3 was setup properly and found solutions before and during the project that allow our web app to re-launch and to be cost effective.
Very solid person and developer. Excellent communication and that is key for me.
Great communication, highly organized, works efficiently without delay. I look forward to working with Mikael again.
I’m Mikael Lirbank, based in Corte Madera, just north of the Golden Gate Bridge in the beautiful San Francisco Bay Area. I live here with my amazing wife and children. In my spare time, I’m a bit of a cycling fanatic, so if you’d ever want to go mountain biking or road cycling in Marin County, hit me up!
I believe everyone deserves to feel safe, respected, and welcome — everywhere, always. That means creating an environment where people can contribute openly, ask questions freely, and be themselves. It also means calling out discrimination and exclusionary behavior — silence isn’t neutrality.
Putting me on a stage or in print? Grab a high-res photo.
I’m an independent consultant in San Francisco. I help teams ship robust, reliable, high-quality AI systems. I care deeply about quality, from AI evals and robust test suites to solid data models and clean architecture.
My main deliverable is peace of mind.
Let’s jump on a call — fifteen minutes is usually enough to know whether I can help.
