Dex

What it is

Dex is an AI recruiter that only places engineers. It covers four role families: AI researchers, machine-learning engineers, software engineers and quant engineers. Engineers sign up for free and hold one voice or text conversation with the Dex agent. The agent asks about their stack, seniority, motivations, dealbreakers and what they want to build next. It then researches companies for them, benchmarks their pay and preps them for interviews. On the employer side, you brief Dex on the role and the team. A proprietary matching engine then picks a small number of candidates who have already said they want this kind of role. Each one is introduced straight to the hiring manager, with no recruiter screen in between. Dex builds the conversation layer on models from Google, Anthropic and OpenAI. The matching itself runs on what the founder calls “old-school machine learning” over the profiles those conversations produce.

The company is based in London and was founded in early 2025 by Paddy Lambros, who advised about 100 European startups on hiring as a talent adviser at Atomico. In April 2026 it raised a $5.3M seed led by Notion Capital, with a16z Speedrun, Concept Ventures and angels from OpenAI joining. That brings total funding to $8.4M. At the raise, Dex reported 15,000+ engineers, 50+ paying companies and about $1.8M in annualized revenue. Named customers include ElevenLabs, Synthesia, Lovable, Granola and Fyxer. Its engineer page now claims 25,000+ engineers. The seed is meant to fund New York and San Francisco offices later in 2026.

Dex is not software you run. It has no seats, no API and no ATS integration. It is an agency with an AI front end, and it bills like one.

Why it shows up in Recruiting stacks

• Candidate intent, not keyword match. Every candidate Dex introduces has already described, in their own words, what they want and what they will not accept. For senior AI and ML engineers that screen is the scarce part. Most of them do not answer cold InMails.
• Very few introductions per role. Dex’s pitch is a handful of matches, not a pipeline to manage. For a hiring manager who spends two hours on every onsite loop, five good candidates are worth more than 50 average ones.
• Senior-weighted pool. Dex says its placements carry a median salary above £150,000 and a floor of £100,000. It also reports 29 days on average from introduction to offer, against 55 days via job boards. These figures are vendor-reported and unaudited.

Pricing

• Contingency, 20–30% of first-year salary, paid only on hire, per Fortune and The Next Web. The rate is not published on Dex’s site. No subscription, no seat fee, no posting fee.
• Real-world cost: at Dex’s reported £150,000 median salary, one hire costs £30,000–£45,000. A US ML engineer on a $250,000 base costs $50,000–$75,000.
• No published replacement guarantee. Fortune, The Next Web and the Dex site don’t mention one. Paraform and many traditional agencies offer a 90-day free replacement.

So Dex does not answer “how do I hire AI engineers without a 20% fee”. Its fee band starts where agency fees start and runs higher. What you pay for is candidate intent and a narrow technical pool, not a lower rate.

Best for

Heads of talent and engineering leaders at London and European AI startups and scale-ups hiring senior AI researchers, ML engineers or quant engineers at low volume. It fits teams where the hiring manager’s time, not the fee, is the binding constraint. It also fits teams whose sourcer has run out of warm leads.

Watch-outs

• The fee is at or above the agency band. Four hires at $60,000 is $240,000, which buys an in-house technical recruiter with money left over. Guard: model your hire count before you sign. If you expect more than about three hires through Dex in a year, hire the recruiter and keep Dex for the hardest research reqs.
• No published guarantee, and the terms are negotiable. Guard: get the percentage, the comp base (base only, or base plus bonus and equity), and a 90-day replacement or refund clause in writing before the first introduction.
• Attribution disputes. Dex’s pool overlaps LinkedIn, referrals and your own applicants. Guard: export your active candidates from your ATS before go-live and agree that anyone already in process within the last 12 months is excluded from the fee.
• Thin independent track record. Dex is under two years old, has 50-odd paying clients and no G2 or Capterra presence. Guard: run it on one or two reqs and track introduction-to-onsite and onsite-to-offer rates in your ATS before you give it more roles.
• US depth is unproven. The pool and the customer list are London-weighted, and US offices were only announced. Guard: ask Dex for the number of candidates in your city and seniority band before you brief a US role.

Alternatives

• Jack & Jill is the fastest-growing entrant in AI talent matching and runs the same conversational matching model, with a much larger pool across more role types. It charges a published 10% of first-year salary and refunds the fee if the hire leaves within 3 months. Pick it when the fee is the constraint or the role is not deep AI or ML.
• Paraform is the leading contingency marketplace, routing your req to specialist human recruiters at 20–25% with a 90-day replacement. Pick it for senior roles outside Dex’s four families, or when you want a named human recruiter working the search.
• LinkedIn Recruiter is the default sourcing channel by market share. Pick it if you have a recruiter with time to run outreach and want to own the pipeline for a flat seat price.
• Juicebox is AI sourcing on a flat subscription. Pick it when you have recruiting capacity and want to cut per-hire cost to near zero.
• Mercor supplies vetted specialist contractors rather than full-time hires. Pick it when you need AI expertise for a project, not a permanent seat.

If none of these fit — you hire more than ten engineers a year in the same families — an in-house technical recruiter using Ashby plus a sourcing tool will beat every contingency model on cost.