Every briefing written about London's AI graduate market in 2026 has anchored to the same cluster of names: DeepMind at Pancras Square, ElevenLabs on Worship Street, Wayve on Goods Way. Those employers are real, their compensation is real, and the career trajectories they offer are well-documented. What none of those briefings have addressed is the employer category that absorbed a quietly significant share of London's computer vision and generative ML graduates in the same cycle — and that is doing so from a geography, a technical domain, and a product trajectory that the frontier-lab narrative framework cannot account for. Synthesia — the AI video generation company whose January 2025 Series D raised $180M at a $2.1B post-money valuation — is the anchor employer of a London AI video cluster that now constitutes the fastest-growing applied AI hiring category outside of coding assistants. The cluster is not a consolation prize for graduates who did not clear DeepMind. It is a distinct career specialisation, and the 2026 graduate market is the first cycle in which it is large enough to be mapped with precision.
What Synthesia Is
Synthesia was founded in 2017 by Victor Riparbelli (CEO), Steffen Tjerrild (COO), Matthias Niessner (CTO), and Lourdes Agapito. The founding team's academic roots are specific and matter for understanding the company's technical trajectory. Niessner was a postdoctoral researcher at Stanford AI Lab before returning to Europe and is currently a Professor of Visual Computing at the Technical University of Munich; Agapito is a Professor of 3D Vision at UCL's Department of Computer Science. UCL's involvement is not incidental or honorary. The research that underpins Synthesia's core avatar and video synthesis technology — dense 3D face reconstruction, neural rendering, photorealistic head reenactment — is continuous with the work that Agapito's group and the UCL computer graphics and vision research group have produced for over a decade. When Synthesia hires computer vision graduates in 2026, it is recruiting into a technical programme that connects directly to one of the UK's strongest academic computer vision lineages.
The product Synthesia has built around that research is now a significant commercial operation. The company's AI video platform — which enables users to generate photorealistic avatar-based video presentations in over 140 languages from a text script, without cameras or studios — serves more than 50,000 companies according to the company's own published figures, including Accenture, Reuters, and Zoom. Revenue crossed $100M ARR in 2024, per reporting by Sifted UK and TechCrunch. The January 2025 Series D, which valued the company at $2.1B, was co-led by NEA and Goldman Sachs, with participation from Accel and existing investor GV. That funding round is the marker that places Synthesia in a different capitalisation tier from the majority of London applied AI companies — above Speechmatics, above Colossyan, above any other name in the AI video cluster.
The company's registered offices in the UK, held under Synthesia Limited (Companies House company number 10865630), are confirmed in filings consistent with a Shoreditch-adjacent operating address in the EC1 and EC2 zone — distinct from the King's Cross AI corridor but within the broader East London AI district that includes ElevenLabs' Worship Street office and the Shoreditch/Tech City cluster. This geographic distinction matters for graduates mapping their first role: Synthesia operates within the same London AI talent ecosystem as the King's Cross employers, but its physical address anchors it to the Shoreditch product-company environment rather than the King's Cross research-lab concentration.
The London AI Video Cluster
Synthesia does not operate in isolation. The AI video generation market in London by 2026 has consolidated around a cluster of employers that is specific enough to constitute a graduate labour market of its own.
Colossyan — founded in 2021 by Hungarian entrepreneurs, headquartered in London, and operating from a pre-Series-B position — is Synthesia's most direct UK peer in the AI presenter and video generation category. The company raised $22M in a Series A in 2023 (per Crunchbase and press reporting) and has maintained a London engineering team of approximately 30 to 40 people. Colossyan's graduate ML engineering band sits at £55K–£68K, below Synthesia's range but consistent with a company at a different capitalisation stage. For a computer vision graduate who wants early equity and a small-team environment, Colossyan is the second-tier entry point in the same technical domain.
HeyGen — the US-headquartered AI video generation company that competes directly with Synthesia on avatar and translation video products — opened a London sales and partnerships office in 2024, with engineering functions following in 2025. HeyGen's UK presence is primarily commercial rather than research-engineering at this stage, but the company has begun hiring UK-based ML engineers focused on the European deployment of its video synthesis pipeline. The HeyGen UK engineering band for 2026 new-grad roles, per ENTRA's recruiter survey, sits at £60K–£72K.
Hour One — the Israeli-founded AI video company that has maintained a UK commercial presence since 2022 — is a smaller name in the same cluster, relevant primarily to graduates interested in enterprise video production tooling and the enterprise sales-engineering hybrid role that AI video companies have created to bridge product and customer deployment. Hour One's UK ML engineering presence is limited; its London headcount is weighted toward solutions engineering and enterprise customer success functions.
Waymark, the US-based AI video advertising company with a UK client base, does not maintain a significant London engineering presence in 2026. It appears in the cluster map for completeness but is not a graduate engineering hiring destination in the current cycle.
The operative cluster for computer vision and generative ML graduates in 2026 is therefore Synthesia — dominant by capitalisation, technical depth, and graduate hiring volume — with Colossyan and HeyGen UK as secondary options at different risk-return profiles. The cluster is geographically concentrated in the Shoreditch-to-City corridor, distinct from but accessible to the King's Cross concentration. This is not ElevenLabs voice-AI territory — the technical problems are fundamentally different (2D/3D neural rendering versus audio synthesis), the academic lineage is different (UCL computer vision versus Cambridge speech and language), and the product surface is different (enterprise B2B video versus consumer and developer voice). The two clusters are complementary attractors for different graduate specialisations within the broader London applied AI market.
What Graduates Actually Build at Synthesia
Synthesia's technical stack is a graduate employer's asset because the problems are well-defined, production-constrained, and immediately deployable — qualities that frontier research labs, by design, do not optimise for. The company's core engineering work spans four domains that map directly onto graduate specialisms available at UCL, Imperial, and Edinburgh.
The first is neural rendering and avatar synthesis. The foundational technical challenge in AI video generation — generating a photorealistic human presenter from arbitrary text input — requires end-to-end pipelines that span face detection, 3D facial geometry estimation, neural texture synthesis, and photorealistic rendering. Graduates from UCL's MSc Machine Learning and MSc Computer Science programmes, whose curricula include computer vision and 3D vision modules directly descended from Agapito's research group, arrive with the conceptual vocabulary to contribute to this pipeline from day one. The graduate ML Engineer role in Synthesia's avatar research team — the most technically demanding entry point in the company's 2026 intake — requires comfort with neural rendering frameworks, differentiable rendering, and the PyTorch-based training infrastructure that Synthesia uses for its avatar model development.
The second is multilingual video synthesis and lip synchronisation. Synthesia's 140-language product requires model architectures that synchronise generated speech audio with avatar lip and jaw motion across phoneme systems that differ substantially between language families. The problem sits at the intersection of audio processing, computer vision, and sequence modelling — and it is the domain in which Imperial MSc AI graduates with audio and vision coursework are finding their way into Synthesia's engineering team. Three Imperial MSc AI graduates from the 2024–25 cohort placed into Synthesia roles through the MSc industry research project conversion pipeline, per ENTRA's understanding of Imperial's placement records.
The third is video infrastructure and rendering at scale. Synthesia generates millions of avatar videos per month for its enterprise customer base. The systems engineering required to scale a neural rendering pipeline — GPU cluster management, model serving optimisation, latency reduction for real-time preview modes — is the domain of the infrastructure and platform engineering roles in the company's 2026 intake. These roles pay at the lower end of the graduate band (£65K–£70K base) but provide production-scale ML systems experience that compounds in value quickly for graduates who intend to move into senior ML platform engineering.
The fourth is generative model research for synthetic media quality. Synthesia's competitive position depends on avatar fidelity improvements that require ongoing research investment. The company's research team — smaller than the engineering teams but influential in shaping the product's technical trajectory — is the most demanding intake and the best-compensated at the graduate level.
Graduate compensation at Synthesia in 2026 runs at £65K–£85K base (~$82K–$108K) across the ML Engineer and Research Engineer intake, per ENTRA's Q1 2026 recruiter survey and two candidate-side conversations. The median confirmed new-grad base sits at approximately £72K (~$91K). Equity is structured through EMI options struck at the January 2025 Series D strike price, with a standard two-year cliff and four-year vest — and at a $2.1B post-money valuation with a clear revenue trajectory toward an eventual liquidity event, the grant-date EMI value for a 2026 new-grad represents meaningful upside. First-year total compensation including grant-date EMI value sits in the range of £90K–£115K (~$114K–$146K) for the median ML Engineer hire. Synthesia holds a Skilled Worker sponsor licence confirmed on the Home Office Tier 2 register; the new-grad band clears the £38,700 Skilled Worker salary floor by a factor approaching two at every level, and international graduates from UCL's large non-EU cohort — predominantly students on Tier 4 visas from India, China, and the US — are routinely sponsored through the Skilled Worker route. The High Potential Individual visa route is available to qualifying graduates from QS Top 50 institutions, providing an additional entry channel for international candidates who prefer employer-independent labour market access before accepting sponsorship.
Applied AI Video vs Pure Research: The Career Fork
The graduate choosing between Synthesia and a frontier research lab in 2026 is not choosing between a good option and a lesser one. They are choosing between two distinct career geometries.
The frontier-lab track — DeepMind Research Scientist, ElevenLabs ML Research Engineer — optimises for publication throughput, theoretical depth, and the reputational capital that comes from working on problems at the boundary of what is understood. The timeline to impact is long. The metrics are academic. The peer group is the global research community.
The Synthesia track optimises for production impact at scale, interdisciplinary breadth (vision, audio, 3D geometry, systems engineering in a single pipeline), and velocity — a model architecture change proposed on Monday is testable against real user video generation workloads by Thursday. The timeline to impact is measured in weeks. The metrics are product quality and latency. The peer group is the London computer vision and generative AI engineering community, which by 2026 is dense enough to constitute a genuine intellectual environment.
The career-development case for the Synthesia track is strongest for graduates whose doctoral or master's research was applied computer vision — neural rendering, 3D face reconstruction, video synthesis — rather than theoretical ML. For those graduates, the DeepMind research track offers theoretical depth at the cost of deployed impact continuity; Synthesia offers deployed impact continuity at the cost of the publication credential. The two are not equally fungible. A computer vision PhD who spent four years on photorealistic face reenactment is not a natural fit for DeepMind's reinforcement learning or language model teams; they are a natural fit for Synthesia's avatar research team, where their doctoral problems are the company's product problems.
The longer-horizon question is equity. Synthesia's $2.1B post-money valuation and £100M+ ARR position it as a credible IPO candidate in the 2026–28 window. The EMI options held by a 2026 new-grad will vest into a public-market instrument on that timeline — a pre-IPO equity narrative that Wayve and ElevenLabs are also selling, but which Synthesia can anchor to a profitable revenue trajectory in a way that pre-revenue lab companies cannot. The risk profile is different from a frontier lab RSU: higher if Synthesia achieves a full-valuation IPO, lower if the market re-rates the synthetic media sector. But for a UCL or Imperial computer vision graduate who understands the product and the sector, it is not an uninformed risk.
How to Get In
Synthesia's 2026 graduate hiring process reflects its academic founding — it is rigorous in technical assessment and specifically calibrated to detect the computer vision and generative model depth that the company needs, rather than deploying the generic ML algorithm coding challenge that less technically distinctive employers use.
The process runs in three stages. The first is a portfolio and code review: candidates submit a GitHub repository and, where applicable, a thesis or dissertation relevant to computer vision, neural rendering, or video synthesis. The quality of the code and the relevance of the academic work are the primary filter. Candidates from UCL Computer Science and Imperial MSc AI who conducted their research projects in Synthesia's partner programme bypass this stage through the conversion pathway.
The second stage is a technical interview focused on computer vision fundamentals and applied problem-solving: candidates are asked to reason about neural rendering pipeline design, failure modes in face detection and landmark estimation, and the latency-quality tradeoffs in real-time avatar generation. The assessment is oral and exploratory rather than whiteboard-coding in character — a format that selects for candidates who can articulate what they understand and what they do not, rather than those who have memorised standard algorithm implementations.
The third stage is a system design interview specific to Synthesia's production context: how would you design a multilingual video synthesis pipeline that scales to ten million monthly renders while maintaining sub-five-second latency for enterprise customers? The relevant preparation is not LeetCode. It is the ML Systems Design literature, specifically the work on serving-time optimisation for large vision models, combined with a clear understanding of the specific tradeoffs between model fidelity and inference speed in the neural rendering domain.
The graduate profiles that Synthesia's 2026 intake is selecting for, per two people familiar with the company's hiring criteria: UCL Computer Science or MSc Machine Learning graduates with coursework or research in 3D vision, neural rendering, or generative models; Imperial MSc AI graduates who conducted industry research projects in computer vision; Edinburgh Informatics graduates with a computer vision and graphics specialism; and international candidates on the HPI or Skilled Worker visa route from ETH Zurich, TU Munich, or comparable institutions where the academic lineage connects to the same neural rendering and 3D vision research tradition that produced Synthesia's founding team.
The Cluster Nobody Covered
The London AI video cluster — Synthesia at its centre, Colossyan and HeyGen UK at its edges, founded on UCL computer vision research and capitalised to a point where the graduate career proposition is structurally serious — absorbed a meaningful share of London's computer vision graduate class of 2026 without appearing once in the briefings that defined the cycle's editorial narrative. That is not because the cluster is small. It is because the frontier-lab frame, which anchors the narrative to DeepMind, ElevenLabs, and Wayve, does not have a natural slot for an applied AI video company with £100M+ ARR and a photorealistic avatar product used by 50,000 enterprises. Synthesia is the employer that fills the gap between the research lab and the consumer app — and in 2026, it is building its graduate bench.
Compensation figures sourced from ENTRA's Q1 2026 recruiter survey (nine London ML agencies) and candidate-side conversations. Synthesia Limited Companies House company number 10865630; registered office and filing data confirmed as of May 2026. Synthesia revenue and customer figures per company-published materials and Sifted UK reporting. January 2025 Series D terms ($180M raised, $2.1B post-money valuation, co-led by NEA and Goldman Sachs) per CNBC and TechCrunch reporting, January 2025. Colossyan Series A funding per Crunchbase. Imperial MSc AI placement figures sourced from ENTRA's understanding of departmental records; figures are estimates and have not been confirmed by Imperial College London. Lourdes Agapito position at UCL Department of Computer Science confirmed via UCL faculty directory. Skilled Worker sponsor status for Synthesia confirmed via Home Office Tier 2 register, May 2026. Synthesia declined to comment on specific graduate compensation or intake volumes.
For the broader London AI graduate destination map, see Where UK AI Graduates Actually Land in 2026. For the Imperial College pipeline that feeds Synthesia's applied AI track, see Imperial College AI Graduates Feed London's AI Corridor in 2026. For the King's Cross voice AI cluster, see ElevenLabs vs DeepMind: How Cambridge ML PhDs Are Choosing in 2026.