get hired at
Deepslate
European speech-to-speech voice AI models
- β
- 4JVACASIV+1
- Berlin
- AI
Open roles at Deepslate
4 positions we're tracking, re-checked daily
- apply β
Enterprise Account Executive (DACH, regulated industries)
berlin, germanyMid-levelfirst seen today - apply β
Solutions / Forward Deployed Engineer (voice agents, SIP/telephony)
berlin, germanyMid-levelfirst seen today - apply β
Speech ML Research Engineer
berlin, germanyMid-levelfirst seen today - apply β
Speech Data Lead (European languages and dialects)
berlin, germanySeniorfirst seen today
Know when Deepslate is hiring before anyone else
A role stays uncontested for about four days. Here's the window, and where we put you in it.
From $5.99/week, cancel any time.
- live
livewatching Deepslate0 applicants
- +2h+2hrole spotted & verified1
- +4h+4hyou get the alert1
- day 3day 3you've applied~8
- day 7+day 7+hits the job boards250+
What Deepslate is building
Deepslate trains its own speech-to-speech model, Opal, that takes audio in and returns audio out with no text transcript in the middle. The standard voice-agent stack chains three systems (speech-to-text, an LLM, text-to-speech), and every hop adds latency and throws away information: tone, emphasis, hesitation, dialect. Deepslate's architecture is three self-trained parts, a speech encoder, a reasoning core built on an open-weights frontier language model, and a speech decoder, trained end to end. The numbers they lead with: 440ms time-to-first-audio on the Artificial Analysis benchmark (fastest listed as of September 2026), about 250ms from end of sentence in their own measurement, around 85% on Big Bench Audio, and the best error rate for European languages on CoVoST2. Opal covers 27 languages. Tech Funding News notes it is the only European model among the six speech-to-speech systems in those benchmarks. The product is developer-facing: WebRTC, SIP and WebSocket endpoints, Python and Node SDKs, and plugins for LiveKit and Pipecat. It runs on Deutsche Telekom's German cloud, is ISO 27001 certified, and can be self-hosted or deployed air-gapped. Pricing is aggressive: 10 cents a minute list, 2 cents for early builders. Customers in production include insurers, contact centres and platforms, though none are named.
Why this matters
Voice is where the cascaded pipeline hurts most. A 2 to 3 second pause is the difference between a call that feels like a conversation and one where the customer starts talking over the bot, and the founders know this first-hand: their 2023 AI meeting assistant died on exactly that latency problem. For European insurers, banks and public bodies, the other blocker is where the audio goes. Paesler's line is that data sovereignty is a prerequisite, not a nice-to-have, and Deepslate discloses where compute happens. A model that is both the fastest on a public benchmark and deployable on German infrastructure or on-prem removes the two objections that keep regulated European call centres on legacy IVR.
Investors: 42CAP (Julian von Fischer), Alstin Capital (Andreas Schenk), SIVentures, Angels: undisclosed business angels
Hiring outlook
Implied but specific. The company says the round will grow its sales and marketing teams and fund model training and European data collection. No careers page or live job posts were found on deepslate.eu.
Working at Deepslate
Deepslate is a AI company based in Berlin, Germany. For a AI company this size, the reality is opportunity to shape your role based on the company stage.
Most Deepslate jobs are based in Berlin, Germany.
How to actually get hired at Deepslate
Why applying the normal way doesn't work
Deepslate uses an ATS, but referrals still come first. Cold applications aren't ignored β they're just behind referrals, sourced candidates, and recruiter picks.
Who to contact at Deepslate
- Who decides
- the hiring manager
- Best channel
- LinkedIn or direct email
What to show them
Wire Opal into a real SIP trunk through the LiveKit or Pipecat plugin and build a German-language insurance first-notice-of-loss agent. Record a call, measure end-to-end latency at the phone (not the API), and write up where the milliseconds go: network, jitter buffer, model, barge-in handling.
A cold email that works at Deepslate
What Deepslate screens for
42CAP's Julian von Fischer cited the founders' technical depth; Alstin's Andreas Schenk pointed to their history of scaling latency-critical systems before Deepslate. That is a bet on a small team that has already beaten OpenAI and Google on a public latency benchmark, now being funded to sell. The hires that matter next are commercial: people who can walk an insurer's procurement team through on-prem deployment and data residency. With a team under 20 and no public job board, the window is a direct note to the founders that shows you understand regulated European voice deployments.
Don't send a generic CV to Deepslate. Mirror the job posting's language to get past automated screening.
Don't make these mistakes
Calling yourself 'passionate about conversational AI'. Paskal and Jan started programming at 11 and built the first voice chat for Minecraft while running two of Europe's largest Minecraft servers; they've been shipping latency-critical real-time systems for millions of users since they were teenagers. Also don't pitch a cascaded STT-LLM-TTS demo as your proof of work. Their entire thesis is that the cascade is the problem.
Mistakes that kill Deepslate applications
At people, a copy-paste CV is immediately obvious. It's an instant no.
Lead with their problem, not your ambition. Deepslate is focused on Voice is where the cascaded pipeline hurts most β show you understand that.
Don't apply and wait. The median AI application gets no response ever. One follow-up at day five roughly doubles reply rates.
Applying to Deepslate? Get the contact, not the form.
The Deepslate interview process
We don't yet have verified candidate reports for Deepslate. What follows is the typical process for a -person AI company, so treat it as a model, not confirmed detail.
Interview stages
- 1
Recruiter Screen
Phone or video Β· 30 min
What it testsBasic qualification and logisticsUsually run byRecruiter or HR - 2
Hiring Manager Interview
Video call Β· 45 min
What it testsRole fit and experience deep-diveUsually run byHiring manager - 3
Technical/Functional Round
Video call Β· 60 min
What it testsSkills assessment and problem-solvingUsually run byTeam members - 4
Final Round
In-person or video Β· 60 min
What it testsCulture fit and cross-functional alignmentUsually run bySenior leadership
Deepslate take-home assignment
Deepslate includes a take-home exercise in their interview process. For AI roles, this typically involves a practical problem that takes 2-4 hours. Focus on clean, working code over premature optimization. They're evaluating how you think and communicate, not just the solution.
Deepslate interview timeline
At 14 days, Deepslate's process is about average than typical for AI (14 days at this size).
Interviewed at Deepslate?
Tell us how it went: stages, questions, timeline. Takes 90 seconds and it's how this page stays accurate for the next person.
Deepslate jobs, frequently asked questions
How many jobs does Deepslate have open?
Deepslate currently has 4 open roles, last verified October 2, 2026.
Does Deepslate hire remotely?
All current Deepslate roles are based in Berlin, Germany.
What roles is Deepslate hiring for?
Deepslate is hiring across Sales, Engineering, Data. The most recent opening is Enterprise Account Executive (DACH, regulated industries).
How do I apply for a job at Deepslate?
Apply directly through the links above, or read our guide on how to actually get hired at Deepslate.
Does Deepslate respond to cold emails?
We're still collecting cold email data for Deepslate.
Who is the hiring manager at Deepslate?
At this size, hiring is usually run by the hiring manager.
How competitive is it to get hired at Deepslate?
Expect 100-250 applicants in the first two weeks for AI roles at this size. Apply within 72 hours for best odds.
How many rounds is the Deepslate interview?
4 stages: Recruiter Screen, Hiring Manager Interview, Technical/Functional Round, Final Round.
Is the Deepslate interview hard?
It concentrates on technical depth and system design rather than abstract puzzles. The stage candidates find hardest is Technical Interview.
Does Deepslate give a take-home task?
Yes, Deepslate includes a take-home assignment.
How long does Deepslate take to get back to you?
Around 14 days across the full process.
What should I prepare for the Deepslate interview?
Study technical depth and system design. At this size (), they care about self-sufficiency over textbook knowledge.
Where is Deepslate based?
Deepslate is headquartered in Berlin, Germany.
Get Deepslate roles before they're posted
A role stays uncontested for about four days. Here's the window, and where we put you in it.
From $5.99/week, cancel any time.
- live
livewatching Deepslate0 applicants
- +2h+2hrole spotted & verified1
- +4h+4hyou get the alert1
- day 3day 3you've applied~8
- day 7+day 7+hits the job boards250+
Related
More AI companies hiring
Other startups we track in the same space.
