Skip to content

The Launchpad Build Cohort

Build a real, deployed AI system in eight weeks. Something completely yours.

For career-changers, self-taught practitioners, and aspiring AI Engineers who want to level up and land a $300k+ role.

If you can already code but have a graveyard of unfinished projects in GitHub, this is for you.

Over eight weeks, up to six people will choose, scope, build, and deploy one real AI engineering project, with me reviewing the technical work every week and making sure the project actually gets finished.

Apply for the October cohort

Takes about 10 minutes.

Learn more about the program ↓
Marina Wyss
  • Six people
  • ·Eight weeks
  • ·October 1 – November 19, 2026
  • ·Thursdays 4:30–6:00pm PT
  • ·Applications close September 18, or when the spots are gone
  • Former Senior Applied Scientist at Amazon/Twitch
  • ·Nearly 400 people coached since 2022
  • ·4.95/5 across 200+ sessions
  • ·150k+ subscribers on YouTube

You don’t need another tutorial. You need to finish a real project.

If you’ve been trying to break into AI engineering, you’ve probably heard the same advice over and over: build projects.

So you do. You come up with an idea, get the first version working, and for a few days everything is going smoothly. Then the scope starts growing, you realize you need all these new components, and pretty quickly the project starts feeling overwhelming.

Since nobody is waiting for the project, nobody notices if you skip a week, and there’s nobody around to tell you whether you're on the right track. So you stop for a few days, which turns into a few weeks, and eventually you give up on it altogether. :(

And with AI, you can get a working demo super fast, which is great, but it also means you can build a surprising amount of stuff without really understanding the decisions behind it. That’s fine until an interviewer asks you why you chose a particular database, how you’d change the system if usage went up 20x, or what you’d do when the model starts giving bad answers.

Anyone can vibe-code a demo now. The useful skill is being able to build something end-to-end that solves a real problem, and then be able to explain why it works.

The Launchpad Build Cohort is designed around that.

Why you'll finish the project this time

Most people who start a legit portfolio project never finish it. You will, because you'll have a deadline, weekly code review, and someone checking on you.

The program is designed to make sure you not only design an amazing project, but actually finish it. You'll have:

  • Structured deliverables and a deadline every week
  • A small group for ideas and accountability
  • Code review from me
  • Access to me via Slack for questions any day of the week
  • Accountability from me throughout the process

What you'll leave with

At the end of the cohort, you’ll have a completely custom, deployed AI system that you chose, scoped, and built yourself. You’ll have a demo you’ve rehearsed, a README that explains the project properly, and an interview prep document built from your own decisions during the project.

By the end you'll be able to show somebody the project, walk them through how it works, and explain why you built it that way when they start asking harder questions.

If you’re aiming for one of the AI engineering roles paying $300k+, that matters a lot more than having another generic project title on your resume.

We’ll use AI coding tools throughout, but we'll use them the right way: to help speed up your process without outsourcing your thinking.

Who this is for (and who it isn't)

Who this is for

If you can already:

  • Write a Python script that calls an API and does something with the response,
  • Use git, and push to GitHub, and
  • Say roughly what RAG is, what an agent does, and what MCP is (at the level of having read about them, not having built with them)…

Then you're ready for this program.

You do not need to be an expert in any of that. You don't need to have deployed anything, and you don't need to be a software engineer with 10+ years of experience. Knowing what the pieces are is enough. Using them properly is what this program is for!

And you don't have to be job hunting. About a third of the people I work with aren't: they’re making an internal move at their jobs, starting to freelance, or building their own product. This all works for this program, too.

Who this isn't for

  • Absolute beginners. If you haven't written code outside a tutorial, this will be a big jump. Consider joining the AI/ML Career Launchpad to get the foundations before the next cohort.
  • Anyone who doesn't have capacity for 10-ish hours a week for eight weeks.
  • Anyone who wants to be handed a project spec. The entire point is that you choose what you build, scope it, and deliver. I'll be there to support you every step of the way, but you need to be able to take initiative.
  • Anyone who wants AI to think for them. See above.
  • Anyone who cannot make a four-figure investment. If that’s you, check out the AI/ML Career Launchpad for more affordable ongoing guidance.

Why I’m running this

My own route into AI/ML was not especially straightforward. My background was in public policy and small business management before I eventually found my way to data science, then ML, and then applied AI engineering, first at a consulting firm in Berlin, then Coursera, and most recently as a Senior Applied Scientist at Twitch/Amazon.

My non-traditional background is actually pretty important, because I remember what it was like to look at jobs in this field and wonder how I was supposed to prove I could do the work before anybody had hired me to do it. “Build projects” is technically correct advice, but it leaves out a pretty important detail: building a real project by yourself is hard, especially when you don’t yet have the technical skill to even know where to start.

Since 2022, I’ve coached nearly 400 people working on transitioning into AI/ML careers, with a 4.95/5 average across more than 200 1:1 coaching sessions. I also run the AI/ML Career Launchpad, my group coaching program, and I’ve spent a lot of time watching where people stall.

Portfolio projects are one of the biggest places it happens.

So this program is designed to help you get past this really important, challenging step.

What coaching clients say about working with me

Program structure

Weeks 1–2 — Decide what you're building, and write a scoping doc.

At the end of week 2, you'll have a thorough scoping doc and my written feedback on it.

Weeks 3–6 — Build, review, correct, repeat.

Start by building the simplest working version first, then improve it. Every week I read your code and suggest the most important things to fix next, so you never spend a week stuck on the wrong problem.

Weeks 7–8 — Deploy, polish, present.

You'll get final feedback from me in week seven and we'll do a demo day in week eight. You finish the program with a live URL, rehearsed demo, a README, and an interview prep doc.

What a normal week looks like

I’ve designed this to fit around a full-time job, which means the live time is deliberately used efficiently and everything else happens async.

You should expect to spend around 10 hours a week total depending on how ambitious your project is.

Wednesday night
Build log due
Thursday, 4:30–6:00pm PT
Live call, 90 minutes
By the weekend
Your code review comes back from me
Every day
Slack, for getting unstuck

Projects are just the first step

Projects teach you the skills you need, and demonstrate that you can do the job. But they're just the first step.

After the cohort, you'll get three months of live coaching from me for free to help you with your job search or ongoing accountability as a Guided member of the AI/ML Career Launchpad (normally $299 per quarter).

Inside the Launchpad you'll get:

  • Weekly calls — live Q&A calls and workshops with me
  • Guest speakers — once or twice a month, like experts on freelancing or mindful AI use
  • Weekly accountability — me keeping you on track
  • Weekly Q&A — for the questions that come up after the cohort ends
  • Role overviews — so you understand the difference between ML Engineer, AI Engineer, and Data Science positions, and which one is right for you
  • A skills checklist and learning tracker app — so you don't spend three months on the wrong skill
  • Vetted external courses — the top ones I'd recommend, so the learning time you do spend is well spent
  • Resume template — the format that's worked for the people I coach
  • Suggested timeline — what to do in what order
  • Presentations on job search topics — finding your path, projects, skill development, networking, interview prep, and more

Do-the-work guarantee

If you do every assignment, attend every meeting or watch the recording, and get to the end of the cohort feeling like it wasn’t worth what you paid, I’ll refund you in full. No questions asked.

Frequently asked questions

That’s fine. Week one is for figuring that out: you’ll bring a few possibilities, we’ll brainstorm together, and I’ll help you choose something useful and a good balance between ambitious and manageable.

If you can write a Python script that calls an API and does something with the response, you’ve used git and pushed to GitHub, and you could roughly explain what RAG, agents, and MCP are, then yes. You don’t need to have built with any of them, deployed anything, used AWS, or trained a model — that’s what the eight weeks are for.

The founding cohort is $2,400, and payment plans are available. This includes the full eight week program and an additional three months of coaching with me for free (see above). This is the lowest the program will ever be priced.

You definitely can! There is abundant free material out there, including some of my own. The bigger issue is, will you? That free material has always been there, so if you’re still stuck it’s something else. That’s what this program is for.

No. A bootcamp teaches you to code, over months, for five figures, usually with a curriculum that's the same for everyone. The Build Cohort is different: there are no lessons or follow-along capstones. It's a way for you to design and build fully custom AI projects.

Courses are good for fundamentals and bad at making you employable, because watching someone else solve a problem doesn't transfer to what you actually need to get hired.

You can absolutely use AI to help build the project. I use AI coding tools all the time myself. But AI is not going to hold you accountable, ask you why you made certain choices, and give you expert advice. Plus if you were going to finish a project on your own, you probably already would have ;)

Yes, it's built for that. The one 90-minute call a week is the only fixed commitment, and it's kept deliberately efficient. Everything else is async.

Calls are recorded, and your code review and Slack access don't depend on attendance.

It might be, but that's really up to you. There's no curriculum in the middle weeks, so the level is set by your project idea and how far you want to take it. If you're a senior engineer moving into AI specifically, this generally works well. If you're already shipping production ML, it won't be a good use of your time (unless you really only want accountability, in which case I'm happy to help).

It works exactly the same way. Roughly a third of the people I work with aren't job hunting, and you'll finish with a deployed system either way.

Under $100 total for cloud services and AI coding credits.

That’s no problem. Consider joining the AI/ML Career Launchpad for support from me at a lower price and slower pace.

There are six seats.

Applications close September 18, or when the cohort fills.

Apply for the October cohort

Takes about 10 minutes.

P.S. This is the founding cohort, which means it’s the smallest I plan to run and the version with the most direct access to me, at the lowest price I'll ever offer.