An engineer who just signed as a research engineer with Mistral, one of the few foundation model labs with more than a billion dollars in funding, has published a detailed playbook on how he got there. According to Hacker News, where the post climbed to 166 points, the whole thing took 18 months of deliberate work. What stands out is that none of it was luck. He worked toward the outcome on purpose, and the method is repeatable.
Here’s the practical guide, with every step preserved in order.
Quick Start
What you’ll learn: how one self-taught engineer moved from a first ML job to a research role at a top lab, and the exact sequence he followed. What you need: an existing entry-level ML or software role, patience for a long game, and a willingness to alternate between deep skill-building and sharp interview prep.
One caveat the author repeats: this is n=1, his own experience. Cherry-pick what fits your situation.
Step 1: Understand tactic vs. strategy
Before anything else, split your effort into two buckets.
- Tactical actions are low effort, high return right now: reading up on a company before an interview, refreshing LeetCode muscle memory, mock interviews, polishing your CV.
- Strategic actions are high effort and can feel pointless in the moment, but compound: learning a technology deeply through a portfolio project, building tenure at a respected org, growing a network, building a personal brand by talking about your work.
Why it matters: strategy builds the career, but tactics decide individual interviews. Skipping LeetCode because it’s annoying can cost you a defining role. You need both.
Step 2: Define the goal
Start inside your current job. The author first booked a career discussion with his direct manager, since more responsibility there would have been the lowest-effort path. When it became clear there was no short or mid-term growth, he knew he had to move.
Why it matters: clear goals are what make strategy possible in the first place. You can’t aim tactics at a target you haven’t named.
Step 3: Mine your network for intel
He reached out to friends and contacts across big tech, startups, scale-ups, FAANG, and big labs. The questions were always similar:
- How do you like your position, and what are the growth trajectories?
- How much work is it?
- What skills, portfolio projects, and credentials do you need to land a role like yours?
This reconnaissance tells you exactly what to build toward.
Step 4: Run your first strategic skill phase
Armed with those answers, he poured hours into LeetCode and bought textbooks to shore up CS fundamentals, most notably distributed systems, data structures, and algorithms. This is the compounding work that later makes interviews winnable.
Step 5: Apply, and expect the first attempt to sting
He sent his first application in August 2024 and did well, but after six rounds he was cut around November. It hurt. Working full time while refreshing ML fundamentals, prepping coding interviews, and doing take-homes left no room to apply more widely.
Warning: trying to job-hunt at this intensity alongside a full-time job is brutal. Something gives.
Step 6: Free up real focus
He resigned, waited out his notice, and went full-time on applications in January 2025. Early results were meager. Many exciting roles never even offered an interview, and in the few he got, nerves led to avoidable mistakes.
Step 7: Get strategic again with structure
Rather than grind harder on applications, he changed the process. He joined the Recurse Center, a cohort-based but self-directed programming retreat. Why it worked on several fronts:
- Freedom to chase ambitious portfolio projects.
- A community doing the same work alongside him.
- A credible new line on his CV.
- A fixed three-month window, short enough to commit to.
The lesson: when tactics stall, zoom back out to strategy.
Why this matters
Foundation model labs are the most competitive employers in tech right now. This account shows the door isn’t only open to PhDs from famous programs. A structured 18-month campaign of alternating strategy and tactics, built on a non-CS background, got someone in.
Your next steps
- Book that internal career talk first. It might be the cheapest win.
- List the exact skills your target roles ask for, then close the biggest gaps deliberately.
- Build one substantial portfolio project, not five shallow ones.
- Treat rejection as data, and switch between strategy and tactics as your results demand.
The full write-up, including his sections on networking, the mindgame, and application touch points, is worth reading at the original source.