Handshake AI

How it works

Handshake AI is Handshake's fellowship programme: it recruits students, graduates and credentialed specialists to train and evaluate large language models for AI labs. Handshake publishes the flow as five steps (apply and verify your identity, get matched to a project, take the training, contribute your expertise, and get paid) and states that no prior AI experience is required.

Matching is done by Handshake's team against its AI lab partners' needs rather than by browsing a board, and its help centre is blunt about the consequence: projects launch on a rolling basis according to partner demand, and there is no fixed timeline for receiving a project invitation. When a project does come up you get an invitation email describing the work, the timeline, the time commitment and the compensation. Some projects ask for additional material such as a portfolio or a slide deck, and onboarding requirements vary from project to project. You set up your payout details only after completing a first task.

Worth knowing

Good:

• Handshake states that rates are posted upfront on each role and do not change once posted
• No prior AI experience is required, and the roles Handshake advertises run from generalist evaluation up to postdoc-level specialisms
• The help centre publishes payout timing in writing and gives fellows a pay processing tracker, so they can see a payment in flight
• Projects carry their own training stage, which Handshake pays for on hourly projects along with task preparation and revisions

Less good:

• Work authorisation is a hard gate on top of the location requirement. Handshake states that F-1 students eligible for CPT or OPT may qualify, that STEM OPT is not supported, and that eligibility has to be confirmed with a school official
• There is no fixed timeline for a first project invitation, and Handshake says so itself: being accepted into the fellowship is not the same as having work
• The binding compensation rules live in the Contractor Agreement and in each project's own terms rather than on any public page, so the detail that actually governs your pay is visible only once you are inside
• Some projects pay per task rather than by the hour, set in the project terms, so both the form and the amount of pay vary by project
• There is very little independent contractor reporting on Handshake AI compared with the older platforms in this market. That is not a criticism of the platform, but it does mean there is less outside evidence to weigh against the company's own account of itself

About this page

Facts above come from Handshake's own AI programme page at joinhandshake.com/ai and from the Handshake AI help centre, last checked 6 September 2026.

Handshake publishes a rate on each individual role rather than a platform-wide range, and the roles shown on its programme page rotate. We have left the typical-pay fields on this page empty rather than assemble a range Handshake has never stated. Read the rate on the role you are actually offered, which Handshake says will not move once posted.

Handshake runs no referral or affiliate programme, so nothing on this page is a paid link and we earn nothing if you apply.

Worker reports are attributed in the sentence that makes the claim wherever they appear above. In Handshake's case there were few to attribute, and we have said so rather than filling the gap.