Google DeepMind's 2026 UK graduate intake runs approximately 65 structured positions across King's Cross and Oxford — a 30 percent increase on 2025 and a near-doubling of the pre-merger 2023 cohort. The structural change is not volume. It is track architecture. The 2026 intake separates, for the first time at this scale, a Research Engineer graduate track from the longstanding Research Scientist graduate path — and that distinction is reshaping which universities DeepMind recruits from, which research areas it is hiring new grads into, and what the two- and three-year career outcomes look like for the people who come through it.
The Research Engineer track has been visible in job postings since Q3 2025 and discussed at Imperial College London recruiting events — but it has not been disaggregated from the broader DeepMind graduate narrative, which defaults to the Research Scientist frame. The two tracks are different jobs, with different university pipelines, different starting packages, and materially different three-year trajectories.
What Are the Two Google DeepMind Graduate Tracks in 2026?
Google DeepMind's 2026 UK graduate intake offers two distinct entry points: the Research Scientist track (targeting PhD holders, ~40 positions, £82K–£88K base) and the Research Engineer track (targeting MEng/MSc graduates, ~25 positions, £75K–£82K base) — and the Research Engineer route, fed primarily by Imperial and UCL, is growing faster.
Google DeepMind's King's Cross campus at Pancras Square runs two graduate entry points. The Research Scientist Graduate Programme — the original, established track — targets candidates completing PhDs with a strong first-author publication record, typically at NeurIPS, ICML, ICLR, or CVPR. It is the track covered in earlier ENTRA reporting on the Cambridge and Oxford pipelines. The 2026 intake for this track runs at approximately 40 UK-anchored positions, with a base in the £82K–£88K range (~$104K–$112K at current rates) and a Google RSU grant valued at £60K–£95K over four years.
The Research Engineer Graduate Programme is newer, larger in the 2026 cycle than most observers realise, and draws from a meaningfully different talent pool. The track targets Masters graduates — MEng, MSc, MRes — whose thesis work sits at the intersection of systems engineering and machine learning: the people who can implement a novel architecture from a paper, benchmark it at scale, and iterate on the infrastructure side of model training. The 2026 Research Engineer intake runs at approximately 25 positions at King's Cross, with a starting band of £75K–£82K base (~$95K–$104K) and an RSU grant in the £40K–£70K range. Total first-year comp for the median Research Engineer new-grad lands at approximately £88K–£105K (~$112K–$133K), below the Research Scientist band but above the ElevenLabs Voice Research Residency and competitive with Wayve's Autonomous Systems Engineering Track on base.
The work is different in character. Research Scientists at DeepMind are expected to generate research ideas, lead projects, and publish. Research Engineers are expected to make those projects computationally tractable — optimising training pipelines, building evaluation infrastructure, and maintaining the systems that let the research operate at the scale of Gemini. The career ceiling is lower in terms of academic output, higher in terms of systems influence. Several senior members of Google DeepMind's Gemini infrastructure team began on the Research Engineer graduate track, per two people familiar with the internal career ladder.
Imperial and UCL: The Two Universities DeepMind Is Recruiting Hardest Now
The Research Engineer track is where the Imperial College London and UCL pipelines become structurally significant. Both universities produce MEng and MSc graduates at volume — Imperial's MEng in Artificial Intelligence graduates roughly 80 students per year, and its MSc AI programme (launched in 2020 in partnership with the Alan Turing Institute) adds a further 120; UCL's MSc in Machine Learning graduates approximately 180 students annually from the Department of Computer Science, with a separate MSc in Computational Statistics and Machine Learning adding 90. (All four cohort figures are per ENTRA analysis of university course registry data; exact cohort sizes subject to annual variation and not independently confirmed by Imperial or UCL.) Combined, Imperial and UCL put approximately 470 ML-adjacent Masters graduates into the London market each year — a volume that dwarfs Cambridge's intake and is increasingly the primary pipeline for DeepMind's Research Engineer track rather than the Research Scientist track.
DeepMind's relationship with Imperial is the more formalised of the two. The lab has maintained an industrial partnership with Imperial's Intelligent Systems and Networks group since 2019, which accelerated substantially after the Google Brain merger created a structural need for Research Engineer-calibre engineers at scale. The partnership, administered through Imperial's I-X Centre for AI in Science on the White City campus, gives DeepMind placement priority for MEng and MSc final-year project students whose work touches training infrastructure, model evaluation, or deployment systems. Per two people familiar with the arrangement, DeepMind's recruiting team attends Imperial's final-year project presentations and has, in three of the last four years, extended Research Engineer graduate offers to students before those students had formally completed the programme. ENTRA's understanding from two people familiar with the 2026 cohort composition suggests at least nine Imperial graduates joined the Research Engineer track, a figure that represents more than a third of the Research Engineer intake.
UCL's pipeline into DeepMind is less structured but equally significant. The Department of Computer Science at UCL's Bloomsbury campus has produced several DeepMind senior researchers — most visibly through the connection between UCL's Gatsby Computational Neuroscience Unit and DeepMind's foundational research on reinforcement learning — and that institutional relationship creates an informal graduate channel that operates below the level of a formal partnership agreement. UCL MSc Machine Learning graduates with strong thesis results in reinforcement learning, generative modelling, or probabilistic inference are routinely referred to DeepMind's recruiting team by UCL supervisors, a process that is more akin to academic recommendation than corporate recruitment. For the 2026 intake, ENTRA's understanding from two people familiar with the 2026 cohort composition suggests UCL contributed approximately seven Research Engineer positions, with a smaller number of Research Scientist positions going to UCL PhD completers.
The combined Imperial-UCL pipeline — an estimated 16 of the 25 Research Engineer positions in the 2026 cohort — represents a geographic and institutional diversification that matters for the King's Cross corridor's long-term talent resilience. The existing coverage of the Oxford and Cambridge pipelines captures the research-scientist tier of DeepMind's graduate machine. The Research Engineer tier, fed primarily by London's own universities, is the less visible but equally essential infrastructure of the combined organisation.
The Research Areas: Where New Grads Actually Work
The 2026 DeepMind graduate intake is allocated across four research areas at King's Cross, in proportions that reflect the combined organisation's current strategic priorities.
Gemini systems receives the largest allocation — approximately 28 of the 65 graduate positions, weighted heavily toward Research Engineers. The work is infrastructure at scale: training pipeline optimisation, evaluation framework development, and the systems tooling that enables Gemini model iterations to proceed at Google's deployment cadence. New graduates on the Gemini systems track are embedded in mixed teams of senior Research Engineers and Research Scientists, working on problems where the research question and the engineering question are entangled — a distinction from pure software engineering roles that DeepMind's recruiting team emphasises in its Imperial and UCL outreach. The comp for Gemini systems Research Engineers sits at the lower end of the band (£75K–£78K base) for 2026 new grads, reflecting the role's more structured onboarding relative to the research-led tracks.
Robotics receives approximately 12 graduate positions, split roughly evenly between Research Scientists and Research Engineers. The DeepMind Robotics lab at King's Cross — expanded substantially following the 2024 merger and now operating in close coordination with Google DeepMind's robotics work in Mountain View — is the newest research area in the 2026 intake. The lab's work on dexterous manipulation and embodied AI has attracted significant internal investment, and the graduate intake reflects a deliberate effort to build out the London robotics bench independently of the Mountain View team. Imperial's robotics and autonomous systems MEng pipeline — specifically the Hamlyn Centre for Robotic Surgery's adjacent robotics research group — is the primary feeder for the London robotics graduate track. New-grad Research Scientists on the robotics track clear £84K–£88K base, reflecting the PhD requirement and the research depth of the work.
Health AI receives approximately 15 graduate positions, concentrated in Research Scientist roles, and draws most heavily from the UCL pipeline. The DeepMind Health team — whose origins in the AlphaFold lineage and NHS data partnerships give it a distinctively London institutional character — recruits graduates with ML training and genuine biological or clinical domain knowledge. UCL's unusual combination of a top-10 ML department and a directly embedded NHS teaching hospital network (UCLH, Royal Free, Great Ormond Street) produces the closest thing to a natural pipeline for this profile. The partnership with UCL's Institute of Health Informatics, formalised in 2022 and expanded in 2025 following the UK AI Action Plan's NHS AI implementation mandate, gives DeepMind access to UCL health informatics research students in their final year — a relationship that functions similarly to the Imperial I-X arrangement for the engineering side.
The remaining 10 positions are distributed across multimodal research, safety and alignment, and a small number of software engineering roles that support the research infrastructure rather than contributing directly to it.
Alumni Outcomes: Where DeepMind Grads Are in Year Two and Three
The two- and three-year career outcome data for DeepMind UK graduate alumni is now large enough to yield meaningful patterns. ENTRA's tracking of LinkedIn career transitions for DeepMind Research Scientist and Research Engineer alumni who joined the King's Cross programme between 2022 and 2024 — a cohort of approximately 85 people with sufficient data — shows three primary trajectories.
The largest group, approximately 52 percent of the tracked cohort, remains at DeepMind or Google DeepMind in a promoted role. Research Scientists who entered at new-grad level have typically progressed to a Staff Research Scientist band by year three — a promotion that carries a base increase to approximately £105K–£115K (~$133K–$146K) and a refreshed RSU grant that materially increases total comp. Research Engineers on the Gemini systems track have progressed to Senior Research Engineer at a slightly faster rate, driven by the acute demand for experienced Gemini infrastructure engineers within the combined organisation. The internal retention rate for the Research Engineer track, at an estimated 58 percent at three years, is higher than the Research Scientist track's 47 percent — a reversal of the historical pattern that saw Research Scientists stay longer, and a reflection of the structural demand for Gemini systems depth.
The second trajectory, representing approximately 28 percent of the tracked cohort, is a move to a UK AI lab or startup — typically Wayve, ElevenLabs, or a Series B-stage company in the King's Cross corridor. The median time-to-departure for this group is 26 months, and the compensation trigger is consistent: a total-comp offer in the range of £130K–£160K (~$165K–$203K) base-plus-equity that a pre-IPO or recently public company can construct through aggressive equity grants. DeepMind's RSU refresh mechanism — which is calibrated to Google Alphabet's broad compensation bands rather than the startup equity market — struggles to match this offer profile without a step-change promotion. The graduates who leave at this juncture are not dissatisfied with DeepMind; they are responding to an offer architecture that King's Cross cannot replicate at year two without a Staff-level promotion.
The third trajectory, approximately 20 percent of the tracked cohort, is a move to a US lab or a US academic position. Research Scientists are more likely to take this path than Research Engineers, and Oxford-pipeline graduates are more likely than Imperial or UCL pipeline graduates — a pattern consistent with the academic orientation of the Oxford Research Scientist cohort discussed in ENTRA's earlier Oxford-Cambridge coverage. The US destinations are primarily Google DeepMind Mountain View (internal transfer, not external departure), followed by Anthropic and academic positions at MIT and Stanford. The internal Mountain View transfer, which accounts for perhaps half of the US-destination group, is not a clean loss to the UK corridor — it is the Hassabis-framed London-research-engine / Mountain View-scale-engine split playing out at the individual career level.
ElevenLabs and Wayve: Competing for the Near-Miss Pool
The graduates who reach DeepMind final-round interviews and do not receive offers — estimated at three to four times the intake number, roughly 200 candidates against 65 positions — constitute the most contested talent pool in the King's Cross corridor. ElevenLabs and Wayve both run structured outreach to this cohort, timed to DeepMind's decision communication window.
ElevenLabs' approach is pitched at the Research Engineer profile specifically. The Voice Research Residency at £75K base with pre-IPO EMI options is structured, per Mati Staniszewski's Q1 2026 LinkedIn posts, around the framing that a voice model problem solved at ElevenLabs (which crossed $500M in annualised recurring revenue in early 2026 and has not published a public user count) ships to its full user base within weeks — a deployment-speed argument that DeepMind's research cadence cannot match. The 2026 Voice Research Residency received approximately 340 applications; the subset of those applicants who also reached DeepMind final-round is estimated by two people familiar with ElevenLabs' recruiting process at approximately 25 to 30 percent of the residency intake. ElevenLabs is not passively absorbing DeepMind near-misses — it is actively recruiting them through conference presence at ICLR and through its Imperial I-X Centre partnership, which overlaps directly with DeepMind's Imperial pipeline.
Wayve's competition for the near-miss pool is targeted at the robotics and vision track specifically. The Autonomous Systems Engineering Track, open at £78K–£90K base with post-Series-C growth shares (~$99K–$114K), is the most comp-competitive alternative for a Research Engineer candidate who interviewed for DeepMind's robotics track and did not receive an offer. Alex Kendall's personal Cambridge engineering network — and Wayve's Goods Way headquarters, which sits a seven-minute walk from DeepMind's Pancras Square campus — gives Wayve a proximity advantage in recruiting this cohort that no Cambridge-headquartered employer can replicate. Wayve's 2026 Autonomous Systems Engineering Track intake of approximately 18 positions drew from a pool in which, per ENTRA's recruiter tracking, at least six candidates had previously reached DeepMind final-round.
The competition for the near-miss pool is the structural mechanism by which DeepMind's graduate programme subsidises the broader King's Cross corridor. The lab's reputation and selectivity attract candidates who would not otherwise apply to London-based AI roles; its rejection rate creates a qualified cohort that ElevenLabs and Wayve can recruit at London-competitive compensation without the brand investment that attracted the candidates in the first place. The challenger employers benefit — and the corridor's overall talent density benefits — precisely because DeepMind's selectivity guarantees a large, qualified near-miss cohort every year.
What's Next: The 2027 Intake and the Imperial Partnership Expansion
DeepMind's Research Engineer track is on a growth trajectory that its current Imperial and UCL partnership capacity will not fully accommodate by 2027. Internal demand for Gemini systems engineers is growing faster than the 25-position Research Engineer intake can supply, and the 2026 cycle has already surfaced hiring pressure: ENTRA understands that DeepMind extended offers to a small number of Research Engineer candidates outside the formal programme cycle in February and March, a mid-cycle hire pattern that signals demand exceeding planned intake capacity.
The Imperial I-X Centre partnership expansion — which ENTRA understands is in negotiation for a formal multi-year renewal ahead of the 2027 intake cycle — would, if concluded, give DeepMind placement priority for a larger share of Imperial's MEng AI graduating class and expand the joint supervision arrangement to include Imperial's Dyson School of Design Engineering, whose MEng in Design Engineering with AI has produced graduates with an unusual combination of embodied systems thinking and ML training. If that expansion proceeds, Imperial could account for 15 or more Research Engineer placements in the 2027 cohort, up from an estimated 9 in 2026.
UCL's pipeline is expanding through a different mechanism: the UCL-DeepMind Health partnership, augmented by the UK AI Action Plan's NHS AI implementation funding, is expected to add a dedicated clinical AI Research Scientist track in the 2027 intake — a track that would be unique among UK AI lab graduate programmes in combining ML research training with structured NHS clinical placement. If the programme launches on the timeline currently understood, it would be announced in Q3 2026, ahead of the 2027 recruitment cycle.
For Imperial MEng and UCL MSc graduates navigating the 2026 market, the practical read is this: the Research Engineer track at DeepMind King's Cross is a structurally credible first job, not a consolation for the Research Scientist route. The systems work is real, the internal promotion track to Staff Research Engineer is well-established, and the two-year alumni outcome — stay at DeepMind in a promoted role or move to a Wayve or ElevenLabs-equivalent with £130K–£160K total comp — is a better career outcome than any UK fintech or consulting employer at the same entry point can reliably deliver. The Skilled Worker visa sponsor licence is active, the processing window for international candidates runs to eight weeks from certificate of sponsorship, and the Global Talent route via Royal Academy of Engineering endorsement is open to Imperial and UCL MEng graduates with a demonstrable research output — a bar that a final-year project with a conference submission clears in most cases.
The King's Cross graduate machine is not one programme. It is two tracks, four research areas, two primary university pipelines, and a near-miss economy that funds the rest of the corridor. That is the structural story of DeepMind's 2026 intake — and it is the one that the next two cohorts of Imperial and UCL graduates should be reading carefully.
Compensation data sourced from candidate-side conversations and ENTRA's Q1 2026 recruiter survey (nine London ML agencies). Graduate intake figures are ENTRA estimates based on recruiter reporting, LinkedIn signal, and university department contacts; they do not represent official DeepMind or Google figures. DeepMind confirmed the Research Engineer programme exists; it declined to comment on intake size or compensation bands. ElevenLabs and Wayve declined specific comment on near-miss recruiting activity. Imperial College London and UCL declined to comment on graduate placement arrangements. Alumni outcome data sourced from ENTRA's LinkedIn career-transition tracking of 85 DeepMind UK graduate alumni (2022–2024 cohorts); figures are estimates subject to incomplete data coverage.
For the Oxford and Cambridge graduate pipeline, see Oxford vs Cambridge: The 2026 AI Graduate Race. For the fintech vs. AI lab competition, see Fintech's Graduate War With Big AI: London 2026. For where UK AI graduates land across all employers, see Where UK AI Graduates Actually Land in 2026.