An ecosystem that mixes research, business and public service
British artificial intelligence is organised around several hubs: London, Cambridge, Oxford, Edinburgh, Manchester, Bristol and a few other university cities. They host national institutes, laboratories of large technology companies, start-ups, consultancies and government departments that recruit people with data and machine learning profiles.
For a student, the challenge is not a lack of offers but how varied they are. A programme reserved for doctoral researchers does not open to a bachelor's student, a research internship does not carry the same expectations as a junior engineer post, and an apprenticeship means an employment contract. Before writing a letter, you therefore need to know which audience the offer is aimed at.
This information remains general; the organisation's own advert always prevails for conditions and dates.
Who can apply: read the programme's audience first
Three examples show how varied the audiences are.
- The Turing Internship Network, run by the Alan Turing Institute, is for doctoral researchers. It links PhD researchers with projects proposed by partner organisations in industry, the public sector or the voluntary sector. The placements are paid, full time or part time, and recruitment takes place in several rounds a year.
- Google's Student Researcher Program is open to bachelor's, master's and doctoral students, with applications organised by level and by location. Placements last from twelve to twenty-four weeks, full time and in person. The page does not list the locations, so you must check that the United Kingdom is among the options.
- UKRI publishes internship and placement opportunities within public research bodies.
The practical conclusion: explicitly filter for the United Kingdom, your level of study and the period, then check that the topic matches your skills. A poorly matched profile is set aside before it is even read.
Internship, PhD, apprenticeship: which format for which level
In the United Kingdom, several formats exist side by side.
For bachelor's and master's students
Summer internships of a few weeks and industrial placement years built into some degree courses are the most common formats. Company labs also offer research placements to master's students.
For doctoral researchers
Doctoral internship networks let you join an applied project during a break from your thesis, with your supervisor's agreement.
For those looking for an employment contract
A British apprenticeship is a job combined with training. There are routes in data and computing, including at levels equivalent to a bachelor's or master's degree. The entry conditions generally include a residency criterion; the government website and the employer must be consulted to find out whether you are eligible.
If your school or university requires a formal internship agreement, check that a British apprenticeship contract would be accepted as an equivalent before signing anything.
Asking for a useful answer, not just a close one: a test to prepare
Many applied AI projects rest on document retrieval: finding, within a corpus, the passage that really answers a question. An excellent preparation exercise is to evaluate a small prototype of this kind.
First write about twenty questions and define, for each, the passages that really answer it. A text that is lexically close, meaning it shares many words with the question, can still be off topic. Then analyse two families of errors separately: missing results, when the right passage does not appear, and misleading results, when a plausible but wrong passage is placed first.
Keep the specific examples for discussion. At interview, showing three concrete cases of misleading results and explaining how you spotted them impresses more than an overall score with no context.
A worked illustration, with round numbers: out of twenty questions, fourteen get the right passage in the top three results, four find it lower down, two do not find it. Among the fourteen, two are accompanied by a convincing but wrong passage in first position. That detail is more instructive than the simple total.
Skills that teams really look at
The skills expected vary with the post, but some come up often.
- Programming, generally in Python, with good code hygiene and use of a version-control tool.
- Solid foundations in statistics, linear algebra and machine learning.
- The ability to build an evaluation protocol and interpret its limits.
- Reading scientific papers and reproducing a published result.
- Written communication in English: a short note, a clear diagram, a summary for a non-specialist.
- Awareness of personal data, bias and responsible use, framed in the United Kingdom as in Europe by data protection rules.
A clean code repository, a short evaluation report or a commented reproduction is often worth more than a list of courses taken.
Right to work, length of stay and logistics
Since Brexit, citizens of the European Union are subject to British immigration rules. Depending on your level of study, the length of stay and the employer, you may need a visa or a suitable status. Some employers support their recruits, others require a right to work already in place. These rules change: read them on the official website and in the advert before applying.
Research placements lasting several months also mean arranging housing, often expensive in London or Cambridge, suitable health cover and proof of funds. If your institution signs an agreement, ask what it covers in terms of insurance and liability.
Finally, the working language is English: fluency in speaking, in meetings and in technical reading must be shown from the application stage.
Putting together a convincing application
An AI application is judged on evidence, not on intentions.
- Choose three programmes or teams consistent with your level, rather than a dozen at random.
- Read the target team's recent work and mention one specific point in your letter.
- Attach a link to your code or a short report, if the instructions allow it.
- Prepare two or three project stories: problem, method, result, limit.
- Expect the classic trick question: how will you know your model is not making systematic mistakes?
Add an academic referee ready to reply quickly. Recruitment campaigns are short and a late reference can cost you the place.
Common mistakes
- Applying to a programme reserved for doctoral researchers with a bachelor's profile.
- Choosing a period or a location that does not match those in the advert.
- Praising a model without describing either the data or the evaluation method.
- Confusing a high average score with a useful answer for the user.
- Neglecting roles close to AI, such as data engineering or evaluation, often more accessible for a first experience.
- Waiting until the eve of the deadline to ask for a reference letter.
If the programme you are aiming for is out of reach, a realistic alternative is to join an open project, contribute to a public repository or target an internship in a more general data team.
Frequently asked questions
Can a bachelor's student do an AI internship in the UK?
Yes, but not in every programme. Some, such as the Alan Turing Institute's internship network, are reserved for doctoral researchers. Others, such as company research programmes, accept several university levels. The page for each programme states the audience, the location and the period.
Does Google's research internship take place in the UK?
The overview page describes an in-person placement in a company office, with applications organised by level and location, but it does not list the countries. You therefore need to consult the application page, filter for the United Kingdom and check that the post matches your level.
How long does an AI research internship last?
The length varies by programme. A company research placement can last from twelve to twenty-four weeks full time. A doctoral internship may be full time or part time. The exact length appears in the advert and in the agreement or contract.
Do you need a visa to work as an intern?
Citizens of the European Union no longer benefit from free movement to the United Kingdom. Depending on the length, the level and the employer, a visa or a status may be needed. Check the official rules and the advert, then ask the employer what it will cover.
Do apprenticeships exist in AI?
There are apprenticeships in data and computing, sometimes at levels equivalent to a bachelor's or master's degree, with an employment contract and training. Residency and age conditions apply. Check them with the British government website and the employer before applying.
What should a portfolio contain when applying?
Favour one or two well-documented projects: problem, data, method, evaluation and limits. A small document search prototype with examples of successes and failures is more telling than a large number of notebooks. Make the code readable and the result reproducible.