Brian Geisel @bgeisel1

What would you do? A friend asks you to recommend them for a job at your company. But you know they’re not a fit. Not even close. Now you’ve got three options: 1️⃣ Lie to your company and vouch for them. 2️⃣ Lie to your friend and say you’ll help, then quietly hope it…

It takes relentless effort, innovation, and grit to grow a company. Giants don’t start as giants 👇 → Steve Jobs sold his VW bus to start Apple → Jeff Bezos and his parents bet everything to launch Amazon → Larry Page and Sergey Brin built Google while still at Stanford And…

“I sent an email” isn’t progress. What actually happened? 📞 Did you call? 🚶♂️ Did you show up? 🚗 Did you go there yourself? A friend of mine needed a permit. Weeks of emails. Nothing. His mentor? Got in the car. Drove to the office. Asked in person. Left with the permit in…

I’ve been testing this for weeks. You describe your idea, Claude Opus 4 builds it. It writes the logic. Designs the UI. Squashes bugs before you can even spot them. The only thing it won’t do? Pay your water bill. (Although… pair it with the right AI agent and it will.) If…

A $100M robotics contract was on the line. And it was about to be lost… over a memory leak. We were called in to fix front-end latency on a robot controller. 💻 Everyone suspected JavaScript or Angular was to blame. But the real culprit? 🧠 A C++ comms layer quietly bleeding…

Some days, I have to shut the door and just think. Not respond to messages. Not jump into tasks. Not even check on the progress of things. Just think. Because as a CEO, that’s part of the job. Everyone else is moving fast—solving problems, checking boxes. And I respect that.…

When I started Geisel Software, I didn’t write a business plan. I had a customer. A real person with a real problem we could solve. That’s what you need to start: someone whose needs you can legitimately meet. Solve one person’s problem well—then find more like them. That’s…

Early in my engineering career, I just wanted to finish the product. So when leadership kept changing direction, it felt like chaos. Now I know better. You don’t build the whole thing and then figure out how to sell it. You build while learning. You talk to customers. You…

Why hardware/software alignment is your project's secret weapon 🤔 In fast-paced fields like embedded systems, IoT, and AI, the difference between project triumph and tech headaches often comes down to one thing… Hardware/software alignment— before you start the build. Here’s…

I’ve seen too many bugs caused by one thing: magic numbers. x = y + 534 What is 534? Who knows. Now someone changes it to 543 in one place… and chaos ensues. When you name your constants (averageDoorWidth = 534), your code becomes readable. Maintainable. Debuggable. Magic…

Everyone’s racing to build the biggest model. But when you’re solving a focused problem, bigger isn’t better—it’s just slowing you down. Small language models trained on high-quality, domain-specific data can outperform the giants. They’re faster to deploy, easier to audit, and…

I’m so excited to share that I will be a speaker at this year’s RoboBusiness Conference & Expo. Join me in Santa Clara on October 15 - 16 as I discuss how to fast-track robotic and automation with smarter AI training data. You can register for the event here:…

Traditional code? Runs on logic. Machine learning? Runs on vibes ⚡️ Old school programming is if-this-then-that. Clean. Predictable. Like a flowchart. Machine learning? It’s: “this probably means that… so maybe do this other thing.” Messy. Probabilistic. But that’s where the…

Stop encrypting passwords 🔒❌ Because if you can decrypt it… so can someone else. You don’t want that. What you need is a one-way hash—math you can only run *one way*. You can’t reverse it. You can only verify if the math checks out. 🔐 For passwords: Hash it. Salt it.…

Don’t hire T-shaped devs. One area of depth? Cute. Not enough. We got called in for front-end JavaScript on a robot controller 🤖 But the real problem? 💻 Not the UI. 🧠 It was a C++ memory leak—buried deep in the comms layer. Most teams would’ve stopped at the UI. Ours…

Most people think synthetic data is just AI making stuff up 🤖 It’s not. We don’t generate data. We engineer reality. 📸 We build scenes with real physics, lighting, materials—down to the tiniest variable. 🧩 Every object, every condition, meticulously designed to reflect the…

Hiring software engineers for their skills? Congrats—you just hired average. We’ve solved some of the hardest problems in tech. Not by chasing buzzwords— But by hiring for greatness. 💡 The truth? Someone “great” at TypeScript... is probably just average at it. Because most…

Everyone’s obsessed with big A.I. models. But what if the small ones are secretly better? Large language models get all the hype. They train on everything—including the junk. 📚 Pizza recipes with glue? Yep. That too. It’s brute force brilliance… but it’s messy. Meanwhile,…

Why are we generating synthetic data for AI? Here’s what no one tells you: AI models are hungry for data then the real world can provide. Instead of just making stuff up, we generate synthetic data which mirrors reality. 🧬 We engineer our data to mirror real world patterns…

We’ve finally solved the data labeling problem—once and for all. Most teams spend two years just getting to the starting line. 🗂️ One year to collect images. 🔖 Another to label them. And even with AI help? It’s often wrong. Bad labels are worse than none at all. That’s why…