4-Year Fully Funded Studentship - Fast Optical Switching for Scalable and Energy-Efficient AI Data Centres

4-Year Fully Funded Studentship - Fast Optical Switching for Scalable and Energy-Efficient AI Data Centres

Quick Decision

Strong but verify fit. This is a four-year, UCL EPSRC Doctoral Landscape Award (UELA) collaborative PhD in the Electronic and Electrical Engineering (EEE) department. The doctoral work is advertised as experimental photonics plus modelling: develop programmable beam-steering optical circuit switching, improve switching speed and optical performance, study how a switch event affects transceiver recovery, and assess consequences for data-centre networks. The listed partner is HUBER+SUHNER Polatis Ltd, with a planned placement of at least six months at its Cambridge premises. The opportunity is most directly aligned with electrical/electronic engineering, computer engineering, photonics or physics backgrounds, and values analytical ability, programming and practical experimental interest. These are project-level welcome/valuable attributes; the formal academic threshold is on UCL's EEE course page.

The important financial distinction is that the collaborative catalogue lists £0 per year as the additional partner stipend enhancement for this project. UELA still advertises its own enhanced stipend, Home-rate tuition and Research Training Support Grant (RTSG); eligible International candidates are to receive a UCL award for the Home/International fee difference, but the programme caps International offers at 15 overall. UCL's scheme page labels £26,022 both as a 2027/28 summary estimate and, in its detailed paragraph, as a 2026/27 figure. The exact annual award should therefore be confirmed in writing. The application deadline is 8 January 2027 at 12:00 UK time; UCL recommends that applications and references be submitted by 16 December 2026, and two references must arrive by the deadline.

Before investing heavily in an application, check fee status and EEE entry requirements, ask how the four-year UELA award maps onto the generic three-year full-time EEE MPhil/PhD route, and request the year-by-year stipend schedule. At interview, clarify the device architecture, performance measures, Polatis placement and industrial mentoring. The advertised six-month placement is a plan in the catalogue; its dates, duties and contractual terms are not published.

Research Problem and Doctoral Work

What is unsolved and why it matters

AI data centres connect many processors and accelerators, so the network must move large amounts of data as workloads change. The project description argues that electronic switching is a scaling bottleneck and associates it with roughly one-third of total power consumption. That is an advertised motivation, not an independently established project-specific baseline. The key research question is narrower and technically testable: can optical circuit connections be reconfigured quickly enough for changing traffic, while the optical path maintains acceptable performance and transceivers resume reliable communication after each change? The UELA project catalogue describes this scope directly.

Work, methods and plausible outputs

The advertised approach combines programmable beam-steering devices, optical experiments and modelling. The student is expected to design systems aimed at improving switching speed and optical performance, examine the interaction between switching and transceiver recovery, and assess implications for future data-centre networks. At Polatis, the catalogue names a shared optical laboratory and experiments involving programmable fast phase light modulation and liquid-crystal-on-silicon (LCoS) equipment. It also identifies experimental photonics, signal analysis, numerical modelling and industrial research as skills to develop. It does not specify a switch model, port count, workload, performance target or required network simulator.

The plausible engineering trade-offs include switching latency, insertion loss, crosstalk, connection reliability, transceiver reacquisition and the energy cost of control. These are useful questions for shaping a thesis, not a published list of required deliverables. The project could connect device-level measurements to a network model, but the catalogue does not state what scale of validation will be expected or how energy savings will be measured.

Four recent UCL-linked papers help locate the work:

• Liu, Odedeyi and Zervas (2024) present a design for a 5 × 5 non-blocking free-space optical switch using piezo-actuators and beam-steering lenses. The paper estimates 1.55 μs switching latency and 2.16 dB insertion loss per connection under optimal alignment. This is a close conceptual comparison for beam steering and routing, but those are design/simulation figures under stated assumptions, not a validated result for the Polatis equipment or this PhD. UCL Discovery record.
• Zhou, Farwell, Wale and Liu (2025) study a tunable DS-DBR laser under optical injection locking. Their tested setup improves coherence and reduces linewidth while injection locking has limited impact on switching time. This informs the transceiver/coherence side of fast optical switching, but concerns wavelength-tunable lasers rather than the advertised beam-steering switch. UCL Discovery record.
• Shabka, Enrico, Almeida, Parsons and Zervas (2024) report deep-reinforcement-learning-based PID tuning on a commercial electromechanical optical switch. Against the manufacturer's production parameters, average switching speed improved by 22%, and the share of switching events stabilizing within 20 ms rose from 17.5% to 87.5%, without practical deterioration in the other reported metrics. This establishes active UCL–Polatis work on switch control, but its hardware/control process is distinct from the phase-modulator/LCoS work advertised here. UCL Discovery record.
• Zhou and colleagues (2023) demonstrate a 51.2 GBd wavelength-switched data-centre interconnect with physical-layer latency below 46 ns using fast tunable lasers, clock synchronization and hollow-core fibre. It shows why transceiver timing and optical propagation stability matter to an end-to-end system. It is a proof-of-concept wavelength-switching study, not a large beam-steering network demonstration. UCL Discovery record.

These papers support the research rationale and show methods in the surrounding group. They do not prove that this project will reproduce any reported speed, loss or latency.

Field momentum and career relevance

The opportunity sits at the intersection of photonics, optical communications and data-centre infrastructure. The Photonics Leadership Group estimates that UK photonics generated £18.5 billion of output and employed 84,800 people in 2024; its industry figures report 8.2% employment growth from 2022 to 2024. The UK Government's 2026 Digital and Technologies Jobs Plan estimates that around 800,000 digital and technology roles will need filling between 2025 and 2035 across covered sectors and occupations. The government explicitly describes this as an indicative hiring requirement, not a target or forecast of net new jobs. These figures indicate a broad industrial ecosystem and policy interest in advanced connectivity, but they do not measure optical-switching vacancies or forecast this PhD's graduate employment or salary. Skills from the project could be relevant to optical communications, photonics engineering, network systems, test/instrumentation or further research; actual pathways will depend on the student's work and hiring conditions. Photonics Leadership Group industry figures; UK Digital and Technologies Jobs Plan.

School, Programme and Field Context

UCL is the academic host and the institution named for the PhD. Polatis is an industrial partner; the catalogue does not indicate a joint or double degree. The EEE MPhil/PhD course page reports that UCL EEE ranked #2 in the UK for Electronic and Electrical Engineering in The Times Good University Guide 2026. This is a department-level subject ranking, not a ranking of this project or the supervisor.

UELA describes four-year registration and expects the thesis to be submitted within the funded period. The generic EEE course page, however, lists three calendar years full-time for its standard MPhil/PhD route. The public pages do not explain how this UELA four-year project is registered, when transfer/upgrade milestones fall, or whether the course-page duration is simply a generic route. Ask EEE/UELA to confirm the project-specific registration and progression plan before accepting an offer.

UELA projects are expected to start on 1 October 2027 unless an exceptional circumstance changes the date. The selection process has an anonymised shortlist expected to finish in late January 2027, interviews expected in early February and final offers expected around mid-March. An informal UELA studentship selection is followed by a separate UCL admissions application and formal course offer. The Optical Networks Group (ONG) is the advertised research group. It works across optical communications, transceivers and network systems. UCL's wider EEE department provides the academic home; access to particular group facilities should be verified at project level.

Supervisor and Lab Assessment

Dr Filipe M. Ferreira is listed as principal supervisor. UCL's ONG profile describes his background in optical communications, spatial-division-multiplexed transmission and transceiver integration, and identifies him as a UKRI Future Leaders Fellow. His current broader research provides a credible setting for a project on optical switching and data-centre interconnects, while his adjacent work should not be mistaken for a fixed thesis plan. The UCL page for Beyond Exabit Optical Communications describes related work on fibre and transceiver integration. UCL's 2025 universal-fibre research update is further evidence of active data-centre interconnect research at the university, but focuses on fibre compatibility rather than this project's switching system.

The ONG's public research and facilities page describes a programmable optical transmission testbed, transceivers, signal generation and detection, and a recirculating fibre loop. Those are group-level resources. The opportunity-specific catalogue instead names the shared Polatis optical lab and programmable phase-modulation/LCoS experiments. Confirm which UCL and Polatis instruments will be available to this student, who will train the student, and whether the six-month placement entails a fixed period on site.

UCL documents an earlier UCL–Polatis collaboration. This corroborates an existing institutional connection. It does not establish the current project's mentoring split, access arrangements or publication/IP agreement. Polatis's technology-partner page describes optical circuit-switch applications from the company's perspective. The company also reports opening a production site in Poland in 2025 and planning a fivefold capacity increase over two years for AI/hyperscale demand; this is vendor-reported business context, not independent market measurement and not the Cambridge placement site. Company announcement.

Funding, Application and Living Cost

Funding and fees

UELA advertises four years of Home-rate tuition, a maintenance stipend at an enhanced rate that rises each year, and an RTSG for training, project costs, conferences and travel. The project-specific collaborative catalogue lists the partner's additional annual stipend enhancement as £0. Therefore, the advertised UELA enhancement should not be confused with a Polatis-funded top-up.

There is a material inconsistency in the official UELA page: the summary block estimates £26,022 per year for 2027/28, while the detailed funding paragraph calls £26,022 the expected 2026/27 total and says it is £1,500 above a £24,522 UKRI London-weighted rate. UCL's separate four-year studentship budget lists base rates of £23,805 for 2026/27 and £24,522 for 2027/28, with later values estimated and subject to annual confirmation. The baseline table helps identify the conflicting year label but does not establish this award's final enhanced schedule. Do not assume that £26,022 is the stipend for each of four years or calculate a four-year total. Obtain the award-year schedule and confirm the RTSG amount in writing. UELA scheme page; UCL studentship rate guidance.

The EEE course page lists 2027/28 International tuition at £36,100 per year and says the UK fee is to be confirmed. UELA states that eligible International awardees will receive a UCL award to cover the Home/International fee difference, but International offers are capped at 15 across the programme. The project is listed as open to Home and International applicants, but individual fee eligibility is assessed by UCL Admissions. Confirm the project-specific fee award and its terms in the formal offer. EEE course fees.

London and visa budgeting

UCL's indicative cost guide lists postgraduate hall rent at £153–529 per week (a 52-week equivalent of £7,956–27,508; postgraduate contracts are usually 51 weeks), shared private rent at £150–300 per week or £650–1,300 per month, excluding household bills, and £426 per month for selected essentials such as food, course materials, phone, health/wellbeing and student travel. These are guide figures, not a personal budget, and the £426 excludes rent and discretionary expenses. UCL cost guide.

The current GOV.UK Student visa page requires proof of £1,529 per month for up to nine months for study in London, subject to listed exemptions and evidence rules. This is an immigration proof-of-funds threshold, not a forecast of actual costs or an indication that a stipend covers every upfront expense. Recheck the rule before applying for a 2027 visa. GOV.UK: Student visa money requirement.

Application requirements and timeline

The official project record welcomes applicants from electrical or electronic engineering, computer engineering, photonics, physics or a related discipline. Interest in optical communications, strong analytical skills, programming and practical experimentation are described as valuable; the advert says training will support specialist skill development. UCL EEE's generic academic requirement is a relevant UK Master's, or at least an upper-second-class UK Bachelor's degree, or an overseas equivalent. It also lists English Language Level 1 and notes that some international candidates may need ATAS clearance, depending on nationality and visa circumstances.

Applications go through the UELA 2027/28 portal. Candidates may apply for up to three projects and must have two references submitted by the deadline; fewer than two references makes the application ineligible. UCL recommends submission by 16 December 2026 because support is unavailable during the Christmas closure. The application prompts include 300 words each for motivation, suitability and knowledge/skills; 500 words each for one research project's description and outcomes; and an optional 200-word contextual section. The UELA page says not to include identifying personal details in these responses. Shortlisting is anonymised at the initial stage, followed by project interviews and a central panel. Formal UCL admission and official transcripts are still required after an informal studentship selection. UELA applicant guide; UELA application and selection information.

Applicant Fit and Preparation

Required background and skills

The catalogue's degree list is welcoming rather than a complete formal eligibility rule. UCL's EEE academic entry requirement is the formal course threshold; an applicant should check equivalence and any additional department requirements. No public project-specific rule makes prior LCoS, beam-steering or optical-switch experience mandatory.

Preferred experience and suggested preparation

The project explicitly values an interest in optical communications, strong analysis, programming enthusiasm and practical experimentation. Useful evidence could include a research or design project where the applicant measured a physical system, analyzed signal data, modeled a communication link or implemented a control routine. This is preparation advice, not an added eligibility condition. Applicants can read the recent papers above and prepare to discuss the difference between optical switching latency, system-level transceiver recovery and network-level performance without assuming that one metric guarantees another.

The work may be less aligned with applicants seeking a purely theoretical topic or a pre-specified algorithm with fixed benchmark and dataset: the catalogue describes a mixed experimental/modelling project, while leaving the technical architecture and performance targets open. That is a scope observation, not a judgment about any individual applicant.

Next Actions and Questions to Verify

Before applying

1. Use UCL's fee-status guidance and confirm that the project remains open to the applicant's fee category; International offers are limited scheme-wide.
2. Check the formal EEE entry requirement, English evidence and potential ATAS timeline. Prepare two references early; aim for the UCL-recommended 16 December 2026 submission date and do not miss the 8 January 2027 noon UK deadline.
3. Ask UELA/EEE which year the £26,022 figure applies to, what the annual enhanced-stipend schedule is, and whether any project-specific RTSG is assigned. The project catalogue's Polatis top-up is £0/year.
4. Ask the supervisor what experimental skills should be evidenced in the application and whether a technical proposal is expected beyond the UELA questions.

At interview

• Which beam-steering/optical-switch architecture and hardware will be used, and what are the planned baselines for speed, insertion loss, crosstalk and switching reliability?
• How will transceiver recovery and network-level consequences be measured or modelled? Which workload assumptions will be used?
• What is the actual split between UCL and Polatis supervision, and who is the named industrial mentor?
• Is the planned six-month Cambridge placement fixed in timing and duration? What work, training, travel and accommodation arrangements apply?
• Does the student have access to broader ONG transmission-testbed resources in addition to the Polatis shared lab?

Before accepting an offer

• Obtain the formal four-year registration, progression/upgrade and thesis-submission schedule, and reconcile it with the generic three-year full-time course page.
• Confirm the annual stipend, payment terms, fee coverage, International fee-difference award if relevant, RTSG, and any placement-related costs in writing.
• Review the industrial agreement for IP, confidentiality, publication review and access to results/data; these terms are not described in the public catalogue.

Evidence and Sources

Official opportunity, school and application

• UELA 2027-28 Advertised Projects Catalogue
• UELA Collaborative Projects Catalogue
• UCL EPSRC Doctoral Landscape Award Studentships
• UELA 2027-28 Guidance for applicants
• Electronic and Electrical Engineering MPhil/PhD
• UCL Electronic and Electrical Engineering Department

Supervisor, research environment and partnership

• Optical Networks Group: About Us
• Optical Networks Group: Research and Facilities
• Advancing Data Centres Through Optical Switches
• Polatis Technology Partners
• Polatis production-site announcement, 12 June 2025
• Beyond Exabit Optical Communications: Towards Transceiver Integration
• UCL universal-fibre research update

Recent primary research

• Liu, Odedeyi and Zervas (2024), beam-steering free-space optical switch
• Zhou et al. (2025), DS-DBR optical switching and injection locking
• Shabka et al. (2024), production-scale optical-switch control optimisation
• Zhou et al. (2023), low-latency wavelength-switched data-centre interconnect

Funding, living costs and sector context

• UCL cost-of-study guide
• Managing research studentships and stipend baselines
• GOV.UK Student visa money requirement
• UK Photonics Leadership Group: UK photonics output
• UK Digital and Technologies Jobs Plan
• UCL contact and UK location