Can you pinpoint the exact moment when a career you love becomes a prison you built yourself?
Mine happened at 2 AM in a hospital room, holding my four-month-old son while waiting for him to sleep to get an MRI.
Hours earlier, consumed by my supervisor’s “urgent” demands about a paper that would only be published another seven years later, I was frenetically jumping between my laptop and the sofa where my baby lay. In that silly dance of academic pressure, my son had fallen to the ground. Sitting in that hospital, I experienced what Daniel Kahneman might call a System 1 override of System 2 thinking. That visceral, immediate knowing that cuts through all our careful rationalizations. No amount of prestige, no publication in Nature or Science, no promise of my own lab could justify what happened. The academic path I devoted ten years of my life was just ending.
But what surprised me most: I felt no fear about leaving. Only relief. Pure, crystalline relief.
The lab is a small startups in a medieval suit
Let me share something counterintuitive about academic research that most people do not realize: scientific laboratories are actually nimble startups trapped in feudal power structures. Think about it. In my immunology lab, we operated with:
- Teams of 8-10 people (classic startup size)
- Rapid iteration cycles (experiment, fail, pivot, repeat)
- Constant innovation and problem-solving
- International, diverse talent pools
- Meritocratic ideals (in theory)
- Passionate people working insane hours for a mission they believe in
This should be a recipe for thriving, creative work. And sometimes, in those magical moments when you are deep in an experiment, surrounded by brilliant colleagues from around the world, all chasing the same elegant solution to a biological puzzle: it absolutely is. The scientific method itself is inherently agile. As Olivier Sibony points out in “Noise”, good decision-making systems require both the ability to recognize patterns (what scientists excel at) and the humility to acknowledge when those patterns mislead us (what the scientific method enforces through experimentation). We are trained to fail fast, iterate quickly, and let data override ego.
So what goes wrong?
What I would call “the prestige trap”
Nothing in life is as important as you think it is while you are thinking about it.
Academia has weaponized this quote from Daniel Kahneman into an entire culture. In academia, people created a system where every decision gets filtered through a single metric: publication impact factor. It is a beautiful example of what is recognized as both bias and noise. Your entire worth—funding, career progression, peer respect collapses into where you publish. This creates what I now recognize as a cascading series of not so optimal decisions:
- You have invested years in specialized training
- Your identity becomes entangled with your academic title
- You optimize for metrics rather than meaning
- You tolerate increasingly toxic behavior because “that’s how academia works”
- You sacrifice everything else, family, health, financial stability for the possibility of academic success
Recent Nature surveys reveal the outcome: 28% of postdocs report moderate to severe depression, with rates of anxiety and depression six times higher than the general population. That is not a bug in the system. It is the predictable output of a machine designed to extract maximum intellectual labor for minimum security. But the very skills that make you excellent at science make you terrible at recognizing when science is altering you.
The failure advantage
When I started learning to code after leaving my postdoc (Thank you Launch School), something felt very familiar. Not the syntax or the logic structures, but the rhythm of the work itself.
Debug, test, fail. Refactor, test, fail. Research, implement, fail differently.
Then it hit me: I have been training for this my entire academic career. In immunology research, failure is the default state. Your experiments fail 95% of the time. Your antibodies do not bind, your cells die mysteriously, your perfectly designed protocol produces noise instead of signal. You develop what I now think of as “failure fluency”. It is the ability to maintain curiosity and momentum when nothing works as expected. Most people entering tech from traditional backgrounds have to learn this comfort with failure. They have been optimized for getting things right the first time, for following established procedures to predictable outcomes.
But scientists? We have been cooked by failure for years. We have developed what is called a “growth mindset” not through corporate training workshops but through daily necessity. I think this turned out to be my secret weapon. When my code did not compile (aaaah scala), when my data pipeline leaked memory, when my machine learning model produced garbage, I did not panic. I did what I always did: formed hypotheses, designed experiments, gathered data, iterated. The medium had changed from proteins to code, but the “meta-skill” remained identical.
What really defines you
Tell me, what is it you plan to do with your one wild and precious life?
This quote from Mary Oliver’s poem stroke me back in the days. What will I do since I was not “Dr. Faouzi Braza, the immunologist” anymore. And here is something we do not talk enough in career transitions: the identity vertigo is real, and it is profound.
I was still a father, still a Brazilian Jiu-Jitsu practitioner, still someone who loved drawing, playing video games and traveling. Yet for academia I became something worse than a failure. I was irrelevant. To family and friends, I had somehow diminished, as if those ten years of research had evaporated the moment I stopped pipetting. This is the “identity foreclosure“. It happens when we commit to an identity without exploring alternatives. Academia does not just encourage this; it demands it. You are not just doing science; you ARE a scientist. But there is some liberating truth: competences are portable, identity is flexible.
A friend put it perfectly: “Faouzi, you are not a scientist. You are a scholar.” That subtle shift changed everything. A scientist is a job title, confined to laboratories and grant applications. A scholar is a way of being in the world endlessly curious, rigorous in thinking, committed to understanding. You can be a scholar while writing code, building data pipelines, developing policy recommendations, cooking, cleaning or building your own business. In other words, I did not leave science. I took the scientific method with me and decided to apply it to new domains.
The consultancy paradox
My journey through tech revealed a fascinating dichotomy that exists everywhere but becomes stark in consulting: the difference between performing competence and possessing it. At Adaltas, my first tech role as an intern, I watched David Worms and his small team of genuine experts manage massive Kubernetes clusters and cloud infrastructures for major French corporations. They competed successfully against consultancies ten times their size. How? They prioritized craft over theater. I was really impressed but the quality of their expertise. But at my next role I encountered quite the opposite. Despite being surrounded by brilliant people, the business model demanded performance over substance. The CEO, phenomenal at sales, less knowledgeable about software development, had sold contracts that did not include time for testing. I was literally told not to write unit tests for a solution because “testing was not in the contract.” Another red flag came during a Big Pharma placement. My manager’s advice still haunts me: “You cannot say ‘I do not know.’” So there I was, performing about cloud engineering concepts I did not master, watching confusion dawn in the client’s eyes as they realized I was bluffing. That shame I felt was massive. The pressure to appear competent often prevents you from becoming competent.
The objective here is not to shoot on any particular organization, but rather to illustrate how consulting environments can vary. What I discovered is that consultancy outcomes depend heavily on two key factors: the company’s underlying business model and their cultural approach to client relationships. Some consultancies, like Adaltas, build their competitive advantage on deep technical expertise. Their business model requires and rewards genuine competence. Others operate on different value propositions: rapid scaling, broad market coverage, or relationship management, where technical depth MIGHT be secondary to other business priorities. Neither approach is inherently wrong, but they create vastly different working environments and learning opportunities. This diversity in the consulting landscape taught me an important lesson: the industry label matters less than understanding the specific culture and incentive structures you will be working in. What looks like the same role on paper can be completely different experiences depending on whether the organization prioritizes craft mastery, client performance, rapid delivery, or relationship building. For anyone considering consulting, I would recommend digging deep into these cultural factors rather than focusing solely on the technical scope of the work.
When Everything clicked at Sunrise
During my time as a consultant, a Belgium recruiter reached out about something different: a machine learning engineer role at Sunrise, a startup developing AI for sleep apnea diagnosis. The first person I spoke with was Cyril Le Mat, their data and software lead: an engineer with a solid track record of shipping ML models in production. Funny fact. This was the first time I acutally had a real technical interview.
Let that sink in.
After dozens of interviews with companies desperately claiming they needed data engineers and ML specialists, Sunrise was the first to actually test whether I could do the work or let put it differently, whether I had the needed skills to learn, iterate and ship. Everyone else had been satisfied with resume keywords and performative conversations about my “passion for data.” Cyril wanted both. First to know if I could actually write code and think about it. Second to know if I add a guenine passion about building data products. The interview went great even though I was not able to answer all the questions correctly. But most importantly, it set the tone for everything that followed. This was a place where building mattered more than talking about building. When I joined, they were running their diagnostic model on a local computer. A brilliant algorithm trapped in a laptop. Everything needed to be built: cloud infrastructure, data pipelines, model optimization, deployment systems.
This was what I had been searching for without knowing it. The startup energy I enjoyed in academic labs but without the feudal overlords. Fast-paced problem-solving where my ideas mattered, where I could contribute directly to the codebase, where learning was not just tolerated but essential. Every day brought new challenges with Google Cloud, new patterns to implement, new systems to design. Cyril became the mentor and colleague I needed all along. Someone who combined technical excellence with human decency, who could teach without condescending, who saw potential and nurtured it rather than exploiting it. Their entire dev team operated this way. We were building something meaningful together, and everyone’s contribution mattered. For thirtheen months, I thrived. I found my confidence not through external validation but through code that worked, systems that scaled, problems that got solved. This was not imposter syndrome territory. This was genuine competence being built line by line, deployment by deployment.
Then came the day that taught me another crucial lesson about modern careers.
The startup reality check
We have nothing negative to say about your work performance or anything else. The company is shifting.
Out of twenty people, eight of us were let go that day. Including me. The company was pivoting from B2B diagnostic technology to an online sleep clinic after their acquisition of Dreem. A completely valid business decision that just happened to make my role redundant. The conversation with Cyril and HR felt surreal in its kindness. No performance issues, no complaints about my work, no dramatic failure. Just a strategic shift. They were genuinely sorry.
I was genuinely devastated.
What hurt most was not the job loss. I understood that startups pivot, that this is part of the game. What crushed me was that they could not envision me contributing to the new direction. Here I was, someone who had transformed from immunologist to data engineer to ML engineer, and somehow I had become fixed in their minds as just the cloud and machine learning infrastructure guy. The very adaptability that got me here had become invisible.
But despite the deception, Sunrise taught me what I actually wanted from work:
- Technical challenges that matter
- Teams that value competence over theater
- Learning as a core activity, not a guilty pleasure
- The ability to see my impact directly in the code and systems I build
- Mentors who teach through collaboration rather than hierarchy
The layoff hurted precisely because Sunrise had given me all of these things. You cannot mourn the loss of something you never valued. The pain was proof that I found something worth finding, even if I could not keep it.
My corporate interlude at AXA
After being laid off, I felt the familiar pressure that every homeowner with a mortgage knows too well. You know the drill. Bills do not pause for career transitions, and the freedom to grocery shop without mental arithmetic is, for me, the ultimate luxury. I do not care about fancy cars or any other fancy furnitures. I just want to live without the constant hum of financial anxiety. And in Belgium, maintaining that baseline requires a steady salary. So when a contractor offer at AXA Belgium came up, I took it. The team specialized in AI and analytics and filled again with talented people working within a very different organizational rhythm than I experienced at Sunrise. We considered ourselves agile, and compared to the main software teams, we actually were. I remember sitting through one of their presentations about CI/CD implementation. Coming from Sunrise where we deployed multiple times daily, the contrast was striking—here, deployments were measured in days, not hours. It was simply a different world with different priorities.
But AXA offered something I had not experienced in years: stability and breathing room. No one breathing down your neck, no impossible deadlines, no existential threats to the company’s survival. After the intensity of startup life, this was a different kind of professional experience entirely. I noticed some of my colleagues, brilliant data scientists with PhDs and impressive publication records, had found creative ways to maintain their technical edge. They discovered a balance that worked for them: the security of a stable position combined with the intellectual stimulation of side projects. Everyone finds their own equation for professional satisfaction.
This experience taught me something valuable about organizational diversity. Large companies do not innovate the same way startups do, and that is not necessarily a failure: it is a feature. AXA needs to keep their existing business sustainable, serving millions of customers reliably. Their innovation might come from a new process optimization, a marketing campaign refinement, or a regulatory compliance improvement. Not from implementing the latest Python library for shipping the last LLM. These are equally valid forms of progress, just operating on different timescales and risk profiles. For me personally, I discovered that this particular rhythm was not quite right. I found myself missing the immediate feedback loops of startup development, the daily problem-solving that had energized me at Sunrise.
It was not that AXA was wrong. It was perfect for many of my colleagues who thrived in that environment. It simply helped me understand more clearly what kind of work environment brought out my best. And as my contract neared its end, I started looking for the next role with clearer self-knowledge. I interviewed for various tech positions, now better able to articulate what I was seeking: a place where I could tackle unconventional problems that needed creative software solutions, where the pace matched my own internal rhythm.
An interesting detour in a think tank
But things do not always unfold as we expect and sometimes, it is where some interesting chapters begin.
During my job search, a peculiar opportunity caught my eye: a think tank seeking someone with a biologist’s background, a grasp of AI, and a curiosity about European policy. The improbability of that combination made me laugh. It felt like someone had quietly mapped my winding career and designed a role just for its odd contours. The interview with my future director, Velislava Petrova, clicked instantly. She was a scientist turned into a global policy shaper, with experience at the WHO, the UN, and even as a deputy minister. She saw something in me I had not fully named yet: not just the biologist or the engineer, but the bridge-builder. Someone who could move between the languages of science, tech, and policy, helping each side hear the other.
For eight extraordinary months, I experienced what work can feel like when purpose, people, and possibility align. We dove into biosecurity frontiers, mapped the bioeconomy’s potential, stood at the crossroads of AI and biotech. A lot of my work consisting on translating complexity for policymakers navigating uncharted territory. It wasn’t just intellectually alive — it felt necessary. My manager’s leadership was a quiet revolution: she assumed competence, invited exploration, and cultivated the kind of psychological safety where real innovation could take root.
Then, as organizations do, strategy evolved and with it, so did the rhythm of my days.
I began to notice subtle shifts. It wass less science and more lobbying with a lot of upcoming public affairs work. Also the company were multiplying processes. Its structures started tightening and conversations growing more cautious. It was not wrong. It was simply… different. And in that difference, I recognized that the Brussels bubble (as insiders call it) is clearly its own ecosystem, humming with influence and institutional gravity. And to have impact in this bubble you need to embrace their rules. For many of my brilliant colleagues, this was exactly where they thrived: navigating political currents, shaping strategy, finding deep meaning in the dance of policy and power. And I admired that. Truly.
For me, though? I began to feel a familiar cognitive dissonance. Not because the work lacked value: it mattered deeply. But because my own compass was pointing elsewhere. I found myself craving the tangible, the iterative, the hands-on. The direct feedback loops of building, testing, adjusting. The messy, fertile ground where ideas meet implementation, not just interpretation. This was not about the organization. It was about me. About recognizing with kindness, not judgment, that I had outgrown the fit. Where others saw leverage, I felt distance. Where they found fulfillment in high-level strategy, I longed for grounded impact. And it is ok. More than ok I think it is natural.
So I made a decision my younger self would have found unthinkable: I chose to leave. No safety net. No next step neatly mapped. By the time you read this, I will have submitted my resignation, effective September 19th, 2025. Not because the mission failed, it burns brighter than ever. But because I have learned to listen when my inner voice says, “This is no longer where you grow.” Sometimes the bravest thing is not to push through. It is to pause, to honor your own evolution, and to step gently into the next unknown.
Lessons learned
After this journey through academia, consultancies, startups, corporations, and think tanks, I have developed a clear picture of what I expect versus what I would like to avoid. The patterns are surprisingly consistent across domains.
What I need:
The places where I felt most alive, whether in a lab, a startup, or those early, electric months at CFG shared a few quiet constants:
Building > Talkingabout building. I need to see the fingerprints of my work in the world — not through reports, dashboards, or org charts, but in real, tangible outcomes. Where decisions about how to build are made by those doing the building — not those selling it, managing it, or approving it from three floors up.Radical transparency, the real kind. Not the performative “we are transparent!” that masks curated messaging. But the messy, honest kind: Here is how we decided this. Here is where the money goes. Here is what we are actually trying to solve. I have seen how secrecy even well-intentioned breeds confusion, mistrust, and ultimately, dysfunction. Clarity, even when uncomfortable, is kindness.Learning as oxygen, not dessert. The places that drained me treated growth as a luxury. Something to squeeze in between billable hours or performance reviews. The places that energized me? They treated learning as core work. They said, “We do not know what we do not know yet and that is exciting.” Not having all the answers is not a weakness. It is acutally the starting point of real innovation.
The death of artificial metrics:
So many systems I’ve moved through measure the wrong things — and then act surprised when people optimize for them.
- Academia’s impact factors.
- Consultancies’ billable hours.
- Corporations’ quarterly targets.
- Policy circles’ proximity to power.
These are not neutral indicators if they become goals. And once a measure becomes a target, as Goodhart’s Law reminds us, it ceases to be a good measure.
What I’m looking for instead? Simpler, truer questions:
- Does the code work?
- Does it solve a real problem?
- Are we learning?
- Are we building something we are proud of?
These are not metrics you can game or perform: they are either true or they are not.
Small teams, big heart and massive impact:
My most meaningful work has always happened in small, autonomous groups. Whether in a research lab, at Sunrise, or those first months at CFG.
There is a magic in small teams: alignment without bureaucracy, accountability without surveillance, contribution without credit-chasing. You see your impact. You feel the shared momentum. Large organizations try to replicate this with “agile” rituals. However you cannot just manufacture intimacy. You can only protect the conditions where it grows.
Trust over surveillance:
The more an organization tries to measure productivity with timesheets, check-ins, KPIs stacked like Jenga blocks, the less actual productivity I have seen. People do not become more effective. They become better at looking effective.
What I value and seek are environments built on trust:
- Where hiring someone means believing they want to do good work not assuming they will slack off unless watched.
- Where flexibility is not a perk, but a principle: adults trusted to manage their time, energy, and focus.
- Where the question is not “Are you working?” but “What are you working on and how can we help?”
Trust is not naive. It is strategic. And it is a strong foundation on which innovation can be built.
What comes next
As I prepare to leave CFG, I am realizing something simple, yet profound: I miss building stuff.
Not just ideas. Not just strategies. But code. Data pipelines. Tools that solve real problems. The quiet joy of debugging at 10 PM because you are so close. The satisfaction of shipping something that did not exist yesterday and now does, because you and your team iworked hard for that. I know the “perfect” organization does not exist. And I am not looking for perfection. I am looking for conscious imperfection. A place that know they are works-in-progress, but are committed to learning, adapting, and growing together. I want to work somewhere that values:
- Craftsmanship over credentials: where what you build matters more than where you studied or how you dress it up.
- Transparency over hierarchy: where information flows freely, and decisions are made closest to the work.
- Learning over knowing: where “I do not know, let’s find out” is celebrated, not silenced.
- Building over performing: where shipping working solutions beats perfect slide decks.
- Quirks over conformity: where different rhythms, styles, and approaches are seen as assets, not anomalies.
- Sustainable pace over heroic sprints: where rest is not a reward, but a requirement. Where “balance” is nt a slogan. It is an operational reality.
- (And yes. Probably a few things I have not even thought of yet… and will fall in love with when I find them.)
These are not utopian dreams. I have seen glimmers of all of them. In David’s quiet mentorship at Adaltas, in Cyril’s collaborative spark at Sunrise, in those early CFG moment when curiosity led the way, even in academia’s luminous moments when discovery eclipsed everything.
What is different now? I know these aren’t “nice-to-haves.” They are non-negotiables not for survival, but for thriving. I can survive without them. I have.
But survival isn’t the goal anymore. The goal is to grow. To learn. To build with joy, with rigor, with heart.
The goal is to contribute to spaces where excellence and humanity do not compete they collaborate.
Le travail est indispensable au bonheur de l’homme ; il l’élève, il le console ; et peu importe la nature du travail, pourvu qu’il profite à quelqu’un : faire ce qu’on peut, c’est faire ce qu’on doit. Alexandre Dumas Fils.