What does it actually take to build a career in AI in 2026? Not the theory — the real, practical answer for a college student navigating a job market that changes every few weeks?
On March 19, 2026, BlendED hosted a live dialogue bringing together two seasoned industry builders and six student leaders from universities across the UK and Europe. The result was one of the most grounded, honest conversations we have had about what college students should genuinely be doing right now.
Our featured speakers were Robin Singh, Senior Optical Scientist at Apple and MIT PhD graduate, and Piotr Swierczynski, CTO of Nodar, an MIT spin-off working on 3D perception for autonomous driving and robotics. The student leaders — Nam, Alessandro, Dylan, Ella, Ben, and Nadeem — each brought questions from their own communities, ranging from robotics clubs to data science societies and hackathon teams.
Here is what came out of it.
Standing Out Has Become a Logistics Problem — and a Skills Problem
Nam Pham, a master's student in AI at the University of Tokyo currently interning at Audi in autonomous driving, opened the discussion with a pointed question:
What actually moves the needle when applying to companies like Apple today?
Robin was direct. The applicant pool has multiplied — for one position, you now compete with hundreds of people, not ten. That changes everything. His advice had two parts.
On the logistics side: proactive, personalised outreach matters far more than a mass application. A generic resume edited through AI tools is immediately recognisable, and does nothing. What does work is a strong web presence — a personal site, a consistent GitHub history, a Google Scholar profile — something that tells a story of sustained work over time. When Robin reviews a candidate, the first thing he does after reading the resume is search their name. What comes up matters.
On the skills side: stop thinking about one fixed skill set. The ability to learn — quickly, from the right sources — is now more valued than any particular specialisation. Robin put it simply: you should be open, not blinded. Whether it is computer vision, photonics, quantum algorithms, or something else, curiosity and adaptability are what hiring managers are actually looking for.
Piotr added a sharper filter: the candidates who stand out are the ones who show genuine ownership. Not a forked GitHub repo with two changes. Something built from scratch, designed, iterated, and brought to a polished result. Real craftsmanship is visible, and rare.
In Robotics, the Biggest Breakthroughs Will Come from Integration
Alessandro Sica, president of the UAM Robotics Society at the University of Manchester and a final-year mechatronics student, asked:
Where robotics development is actually heading — toward better AI models, better hardware, or something else?
Piotr gave the clearest answer of the session. The constraint is not models and it is not hardware. It is the interface between them. We already have models smarter than our robots can use, and hardware more capable than software can exploit. The engineers who understand both sides — who can work across that intersection — are the ones who will succeed.
The reason is practical: models and hardware are too often designed independently, by different communities with different assumptions, and they do not speak the same language. Robotics also introduces real-world constraints — latency, power, safety — that benchmark performance alone does not account for. The remedy is co-design: hardware and software developed together, with each side shaping the other.
Robin reinforced this with a concrete example. The reason Apple Silicon outperforms competitors is precisely because macOS and Apple's chips are built together. The same principle applies to robotics at every level.
For students, both speakers agreed: being on that hardware-software intersection makes you genuinely harder to replace — by both companies and AI.
Passion Projects Can Become Products — But Only If Someone Else Has the Problem
Dylan Kainth, head of events at KCL's tech society and organiser of Hack London and the UK's largest student tech fair, asked:
How student passion projects can actually become products, businesses, or careers.
Robin emphasised motivation and honesty. Before anything else, ask yourself: is this a hobby, or does it solve a real problem in the world? And if it is the latter, can it attract funding? The current moment is unusually good for prototyping — AI companions can fill skill gaps between hardware and software in ways that compress development cycles significantly.
Piotr gave the most direct advice of the session: before you write a single line of code, find one real person other than yourself with the same problem. Most projects do not fail on execution. They fail because no one needed them in the first place. He described watching a friend spend 18 months building something that no one wanted — a genuinely soul-crushing experience that was entirely avoidable.
The practical sequence he recommended: validate first, build an MVP as fast as you can, get it in front of real users with no shame, and let people pay. Perfection is not the goal at that stage.
The Questions Students Are Not Asking Enough
Ella Tekeli, a final-year arts and sciences student at UCL and head of research at the Data Science Society, asked:
What techniques, applications, or implications they wish more people were thinking about?
On the technical side, Piotr flagged three areas he sees as underexplored: world models (which will attract significant investment beyond Jan LeCun's own work), Bayesian deep learning with proper uncertainty quantification, and continual learning — the ability for models to update over time without drifting. Robin pointed to wearable AI as a hardware-software integration challenge with enormous room to grow, and to the need for numerical LLMs — models trained to reason about quantitative data with measurable confidence.
On implications, both speakers were measured but honest. The labour market disruption is real but slower than the headlines suggest. The biggest near-term effect Piotr sees is not mass unemployment but scope reduction — fewer people doing the same work. He also raised a second-order effect that gets less attention: if companies stop hiring junior engineers today, where do senior engineers come from in ten years?
Robin framed AI's current trajectory the way historians might frame the industrial revolution. Mechanical jobs were replaced by machines, but new roles emerged. This is the same shift, one level up — from manual coding to AI-assisted building. The question is not whether the shift happens, but whether you are positioned to move with it.
Should You Do a Postgraduate Degree? An Honest Answer
Ben Chesworth, founder of the Turing Society and a third-year maths student, asked a question many students avoid asking directly:
Is there any point in graduate school right now, when you could be building things instead?
Robin's answer depended on where you are and what you want. If you feel underprepared for the direction you want to go, a postgraduate degree can provide structure that the internet cannot. There is real value in the discipline of working through a problem systematically over months — a muscle that is harder to build independently in a world full of distraction. But if your goal is a startup or early employment, your undergrad time is enough — use it to build projects and apply.
Piotr added something important: the real output of a PhD is not a thesis. It is meta-skills — the ability to struggle productively in solitude, maintain focus on a single problem for months, and develop genuine work ethic. Those are undervalued, and they compound over time.
His recommendation for most graduates, though: get employed first. Work with the best people you can find. Learn what problems exist and what has already been solved. Spend two or three years just learning, then return to academia if something specific calls you back. Solo projects can happen on the side — but it is very hard to sustain them without income.
Cloud AI or Edge AI? Where to Build for the Future
Nadeem Ahmad, treasurer of the Robotics Society at the University of Manchester — the UK's largest, with over 550 members — closed the student questions with one that combined career strategy and environmental responsibility: given the growing water footprint of data centres, should students be doubling down on cloud AI or pivoting to edge AI?
Piotr's answer was clear. Both are here to stay, but the talent shortage is on the edge side. If technical strength and long-term resilience are your goals, edge AI is a stronger bet — it requires both hardware and software understanding, demands real-world optimisation under constraints, and is significantly harder for LLMs to replace than cloud API work. The cloud will remain important, but most cloud AI jobs will ultimately come down to calling APIs rather than building.
Robin agreed, adding that the current generation of AI infrastructure is essentially a prototype at scale — the next wave will be about efficient models running on edge devices, not just stacking more GPUs. The sustainability concern Nadeem raised is real, and the field has not adequately addressed it. That gap is an opportunity.
On a separate note, Nadeem also asked about ethical engineering — how to ensure that your work, over time, is something you are proud of. Piotr's answer was direct: ask yourself whether, looking back in ten years, what you built made a positive impact. Not just whether it was technically interesting, but whether it was good. Robin added that ethical principles should not change with the technology — they need to be foundational, not adapted to whatever the current opportunity is.
Watch the recording ↓
Passcode: p7^=ks!Q
About BlendED
BlendED is an AI education platform based in Kendall Square, Boston — right next to MIT. We help college students from over 20 countries build a real, tangible profile in AI through project-based learning, MIT-certified coursework, and mentorship from industry experts and researchers.
Our AI+X Learning Plan runs 6 to 12 months and connects students with projects across computer vision, NLP, robotics, finance, biotech, and more. Students who complete the programme leave with a portfolio that admission offices and recruiting teams recognise.
Beyond the programme, BlendED co-founded the AI+X Global Talent Community — a network now spanning partnerships with over 150 universities and 100 student clubs across Brazil, Mexico, South Korea, France, Singapore, South Africa, Taiwan, Japan, the UAE, and more.
Events like this one are what the community makes possible. If today's conversation made you think about what you should be building next, that is exactly what we are here to help with.
👉 Explore the BlendED program catalog | Join our next event | Connect with our team