Who Gets Hired
A 22-person company does not hire like a hundred-person Series B startup. The roles are fewer, the scope wider, the filter sharper — a single mis-hire at this stage reshapes the product trajectory. kapa.ai sits in that narrow window where every engineer touches the retrieval stack that OpenAI, Docker, and Nokia trust in production, and every go-to-market hire carries a quota that funds the next research sprint.
The team splits into two tracks. Engineering shows two active roles: Software Engineer (Full-stack) and Research Engineer, Applied AI. Both require three years' experience. Salary bands run $100K–$150K, as Zero G Talent's board lists, and $100K–$170K respectively, with equity at 0.10%–0.30% for the full-stack role and 0.10%–0.20% for the research role. The stack centers on Python and React, but the work is not generic full-stack development. The research track builds and optimizes the RAG pipeline, including retrieval, reranking, context pruning, and citation grounding, which powers the accuracy claims kapa.ai publishes: 99% on customer benchmarks against competitors scoring 60%, which kapa.ai reported. The full-stack track ships the interfaces and integrations that let Monday.com deflect 98% of tickets, based on kapa.ai's figures, and Raspberry Pi move from A/B test to live in six weeks. CTO Finn wrote low-latency trading systems at Bloomberg before co-founding; that DNA, measurable correctness over clever abstraction, sets the bar for both roles.
On the commercial side, four roles are live: Account Executive (US) at $150K–$300K OTE, which Zero G Talent's job board reports, Account Executive (Europe) at $120K–$200K OTE, as Zero G Talent's data shows, Senior Business Development Representative (US) at $80K–$120K, and GTM Associate (US) at $70K–$120K. A Marketing Lead role ($70K–$110K, 0.03%–0.10% equity) rounds out the non-technical slate. The split reflects a deliberate motion: two enterprise AEs covering US and EU time zones, supported by a senior BDR and a GTM associate who runs outbound and pipeline operations. CEO Emil Sorensen, a former McKinsey engagement manager, still runs sales cycles himself; the hires are expected to operate at that density — technical fluency to demo to CTOs, commercial discipline to close six-figure deals without a dedicated sales engineer layer.
What connects both tracks is not a pedigree filter. The YC page notes the founders met six years ago across two master's degrees, Finance at LSE and Computer Science at Imperial, but the job posts list no degree requirements. The research engineer role asks for "advanced degrees or significant experience in machine learning, LLMs, and information retrieval." The "or" does heavy lifting. A candidate who shipped a production RAG system at a prior startup clears the bar without a PhD. The full-stack role asks for three years building product, not three years at a FAANG. The go-to-market roles index on developer-facing sales cycles, Docker, Grafana, and Mixpanel are named reference customers, because the buyer is technical and the proof is a live demo against their own docs.
The hiring pattern that emerges is applied depth over credential breadth. The company's own blog, which includes posts on pruning 68% of RAG context, indexing images for retrieval, and benchmarking on messy real-world knowledge, functions as a public work sample. Candidates who read those posts and can argue with the trade-offs advance. Candidates who lead with framework names stall. The next section maps what that filter pays.
The Pay Structure
kapa.ai's compensation reflects a Y Combinator-backed team (S23 batch) still early enough for equity to carry real weight but mature enough to post defined cash bands across nine salaried roles. The company's job board lists base salary ranges from $60,000 for a Commercial Associate in Denmark up to $300,000 for a U.S.-based Account Executive, and Zero G Talent's board reports a median posted band of $120,000. Equity grants are tiered by function: engineering and research roles carry 0.10%–0.30%, go-to-market roles 0.01%–0.10%, and the single marketing lead slot 0.03%–0.10%. All figures are quoted in U.S. dollars regardless of location.
| Role | Location / Remote Scope | Base Salary (USD/yr) | Equity Range | Experience Floor |
|---|---|---|---|---|
| Account Executive (US) | New York, NY / Remote (US) | $150,000 – $300,000 | 0.01% – 0.05% | 3+ years |
| Account Executive (Europe) | Denmark / Remote | $120,000 – $200,000 | 0.01% – 0.05% | 3+ years |
| Research Engineer, Applied AI | 20+ EU countries / Remote | $100,000 – $170,000 | 0.10% – 0.20% | 3+ years |
| Software Engineer (Full-stack) | 20+ EU countries / Remote | $100,000 – $150,000 | 0.10% – 0.30% | 3+ years |
| Senior Business Development Rep (US) | New York, NY | $80,000 – $120,000 | — | 1+ years |
| GTM Associate (US) | New York, NY | $70,000 – $120,000 | — | 1+ years |
| Business Development Rep (US) | New York, NY | $70,000 – $110,000 | — | 1+ years |
| Marketing Lead | Denmark | $70,000 – $110,000 | 0.03% – 0.10% | 3+ years |
| Commercial Associate | Denmark | $60,000 – $100,000 | — | 1+ years |
Two patterns stand out. First, the equity spread between technical and commercial tracks is deliberate: a founding full-stack engineer can receive up to 0.30% while a senior U.S. account executive tops out at 0.05%. That gap signals where the company believes leverage lives — retrieval accuracy, latency, and token efficiency are engineering problems that compound. The benchmarks kapa.ai publishes show 5.1-second latency versus 16.8 seconds for alternatives, and 5,100 tokens returned versus 40,800. Second, cash bands for the same title vary by geography but not by remote eligibility; a Research Engineer in Germany or Ukraine sees the same $100K–$170K band, which Zero G Talent found, as one in the U.K., and the full-stack role is explicitly open across 20 European countries plus Egypt, Turkey, and Kazakhstan. The U.S. sales roles cluster in New York with a remote option, while the European sales and marketing roles anchor in Denmark, the company's registered base per YC filings.
Benefits detail is thin in the public postings — no line items for health, retirement, or learning stipends appear on the board. What is explicit: every role lists a minimum experience requirement (one year for early-career GTM, three for IC engineering and senior commercial), and the equity grants are expressed as percentage ranges rather than share counts, standard for a pre-Series A cap table still being allocated. The board's aggregate view (nine salaried roles, $68K–$220K band, $120K median) — where Zero G Talent's figures put the band — puts kapa.ai squarely in the seed-to-Series A European deep-tech compensation tier: above pure pre-seed cash, below late-stage FAANG total comp, with the equity upside tied to a product that already serves OpenAI, Logitech, Monday.com, Nokia, Raspberry Pi, and Nordic Semiconductor at millions of queries per month.
Inside the Hiring Funnel
kapa.ai runs a three-stage funnel that typically resolves in seven days — faster than the ten-day median for B2B productivity startups under 50 people. There is no recruiting team. At 22 employees, every application lands directly with a founder who is also running sales, product, and payroll. The bottleneck isn't an ATS queue; it's being seen at all.
First Contact: 30-Minute Video Call
The first screen is a conversation with a founder or the hiring manager. It tests culture fit and role expectations. Recruiters don't run this; the people who will sign your offer letter do. They're listening for whether you understand what kapa.ai is dealing with right now, scaling an AI-powered knowledge platform across enterprise customers, and what you'd do about it. A generic "I want to work at an AI startup" gets discarded. A tailored answer that references the product's RAG architecture or its PDF-to-markdown conversion challenges signals you've done the homework.
The Deep Dive: 60 Minutes on Your Craft
A technical founder or lead walks through your past projects and problem-solving approach. This isn't a LeetCode session. The team cares about how you operate inside a codebase of moderate complexity, how you reason about trade-offs, and whether you can work autonomously. The GitHub take-home repository for the Research Engineer role makes this explicit: "Understanding how you think about this problem is equally important to us as the actual solutions you come up with." Candidates document their thinking in a markdown file as they go. The task, improving a PDF-to-markdown converter for a RAG system, is a real problem the team faced, packaged as a toy app. Reviewers include [email protected], [email protected], and [email protected]. If you submit a solution, you receive a $300 Amazon gift card for the time investment.
Contradiction note: The company's public profile states "No, kapa.ai does not include a take-home stage," yet the Research Engineer pipeline clearly uses one. The discrepancy likely reflects role-specific variation, the take-home appears targeted at research-heavy positions where evaluating open-ended problem solving matters more than algorithm trivia.
Final Round: 45 Minutes with the Founders
The founding team runs this stage. Candidates report it as the toughest. It tests team fit and transitions into offer discussion.
What Gets You Through
Apply within 72 hours. Roles at this stage typically draw 50–100 applicants in the first two weeks. The single biggest lever you control is speed.
Cold outreach outperforms the application form. Since a founder reads every inbound, a concise, tailored message to [email protected] or [email protected] referencing a specific technical challenge from their blog or GitHub repos gets read. The form goes to the same inbox but sits lower.
Tailor the CV to the job description language. The filters are manual but sharp: relevant experience and demonstrated autonomy. Open with what you'd do about kapa.ai's current growth phase, not what you want.
Follow up on day five. Data from similar-stage B2B companies shows a day-five follow-up can double response rates. Most applications get ghosted because the founder forgot, not because you're unqualified.
Document your reasoning. Whether in the take-home or the deep dive, a well-documented journey (ideas tried, paths rejected, what you'd attempt next) often carries more weight than a flawless end result. The team has said as much in writing.
Two Hubs, One Team
kapa.ai operates from two physical offices, New York and Copenhagen, and hires remotely across the United States and a broad swath of Europe. The company's about page lists "NY · CPH Offices in New York & Copenhagen" and its careers copy states the team is "hiring researchers and engineers, across Copenhagen, New York, or remotely." That dual-hub, distributed model shapes how the 22-person team collaborates and where specific functions concentrate.
The New York office anchors the go-to-market side. Board data shows three roles explicitly tied to New York: Account Executive (US), Senior Business Development Representative (US), and GTM Associate (US). All three are listed as New York, NY, with a remote option for the Account Executive. This clustering suggests the New York site functions as the commercial hub — close to the density of U.S. technical buyers and the investor network that backed the S23 raise.
Copenhagen hosts the other physical office and, by extension, a concentration of the research and engineering talent. The founders, Emil Soerensen and Finn Bauer, are Danish; the company's YC profile and LinkedIn presence both signal Copenhagen as a founding location. The board's Research Engineer role lists the United Kingdom first among its eligible countries, followed by Germany, France, Norway, Denmark, Sweden, Finland, and a long tail of European states — all with a remote option. The Software Engineer posting mirrors that European-first country list. The pattern is clear: Copenhagen is the technical center of gravity, with the office providing a coordination point for a team that also spans the continent.
The remote policy is not an afterthought. Every engineering and research posting includes a remote option across the listed European countries, and the U.S. Account Executive role explicitly offers "Remote (US)." The European Account Executive role lists "DK / Remote" — Denmark or remote. This is a deliberate distributed model, not a pandemic remnant. For a company building retrieval infrastructure that ingests documentation, tickets, wikis, and code from hundreds of customers and answers millions of questions monthly, the ability to hire senior applied-AI researchers in Berlin, Stockholm, or London without relocation friction expands the talent pool well beyond what two small offices could support.
What each location enables follows from that split. New York gives the GTM team proximity to the buyer personas, developer-tool companies, enterprise software vendors, hardware manufacturers, that kapa.ai sells into. The office serves as a base for in-person customer meetings, investor check-ins, and the kind of rapid iteration on sales motion that early-stage companies need. Copenhagen gives the research and engineering team a physical anchor for the deep technical work: building the retrieval benchmarks (the Company Knowledge Bench launched in 2024), optimizing the agentic retrievers that score 0.65 on that bench at roughly five seconds latency versus the earlier figure for grep-based baselines, and maintaining the 200-plus source connectors that auto-sync everything from scanned PDF service manuals to Zendesk tickets and GitHub repos. The office also hosts the founders, keeping product direction and technical strategy in the same room.
The remote layer stitches it together. A Research Engineer in Munich or a Full-stack Engineer in Barcelona ships code into the same monorepo as the Copenhagen desk; an Account Executive in Chicago runs deals alongside the New York team. The board's salary bands, $100K–$170K for research, $100K–$150K for full-stack, $150K–$300K for U.S. sales, are location-agnostic within each region, signaling that output, not office attendance, sets compensation. For candidates, the choice is practical: join a hub if you want daily in-person collaboration with the core team; work remote if you're in a covered country and prefer autonomy. Both paths feed the same product, grounded AI assistants that cite sources and admit "I don't know" rather than hallucinate, and both are evaluated on the same applied-technical bar the hiring process filters for.
The Profile That Fits
The hiring pattern at kapa.ai reveals a clear filter: the company selects for engineers and researchers who can move fluidly between deep technical work and the messy reality of shipping product to other developers. Since its 2023 founding by Finn Bauer and Emil Soerensen, the team has grown to roughly 22 people spread across Europe and the U.S., and every open role, nine as of the latest board data, sits at the intersection of applied ML and production engineering. The Y Combinator backing signals a baseline of technical credibility, but the actual bar is defined by the work itself: building retrieval-augmented generation systems that Docker, Grafana, and Mixpanel trust as the first line of defense on their support queues.
That work demands a specific kind of technical fluency. The board listings for Research Engineer and Software Engineer both anchor in Python and React, but the research engineer role explicitly calls for this requirement. In practice, this means candidates who have wrestled with chunking strategies, embedding drift, and evaluation pipelines, not just those who have fine-tuned a model on a clean dataset. The product is an ingestion and retrieval system that unifies 50-plus sources (documentation sites, PDFs, tickets, community threads, API specs) and keeps them current as the underlying content changes. Engineers who thrive here are the ones who treat that freshness problem as a systems challenge, not a modeling afterthought.
Product obsession shows up in the customer list. When Grafana deploys kapa as a chat interface on public docs, or Mixpanel uses it to deflect support tickets, the quality bar is set by developers who will notice hallucinations immediately. The company's own blog describes "applied lessons from teams building AI agents for technical knowledge", a framing that positions the engineering team as practitioners first, researchers second. Candidates who have shipped LLM features to production users, especially in developer-facing contexts, advance further than those with publication records but no deployment scars.
The collaborative signal is structural. With 22 people across at least 15 countries (the Research Engineer role lists GB, DE, FR, NO, DK, SE, FI, PT, ES, BE, NL, IT, CH, AT, CZ, PL, EE, LV, LT, SK, HU, SI, HR, RU, UA as eligible locations), the team operates remote-first by necessity. The board data shows salary bands that are location-agnostic within broad regions, $100K–$170K for research engineers across that entire European list, which means compensation doesn't anchor to a Bay Area premium. People who thrive in this setup are self-directed communicators who can synchronize across time zones without daily standups becoming a crutch. The GTM Associate and Senior BDR roles based in New York suggest a small but deliberate go-to-market motion that sits close to product; engineers who understand that feedback loop, support tickets becoming training data, product gaps becoming roadmap items, integrate faster.
Pedigree is explicitly secondary. The Talantir analysis of hiring patterns notes the company's requirement, the "or" doing heavy lifting. A PhD from a top lab helps, but a portfolio of shipped RAG systems, contributions to open-source retrieval tooling, or a track record of debugging production embedding pipelines carries equal or greater weight. The main theme of this guide holds: applied technical skill over pedigree.
Where the research goes thin is on direct employee voice. The Indeed and Glassdoor reviews indexed under "Kaap" and "Cape AI" describe a different company, an industrial operation with maintenance technicians and paraprofessionals, not an AI startup. No verified kapa.ai employee reviews appear in the available sources. That absence is itself a signal: at 22 people, the team is small enough that public reviews are sparse, and the culture is still being written by the people currently in the room. Candidates who need a well-documented culture deck before they join will wait a long time. The ones who get through are comfortable evaluating a team by its technical output, its customer list, and the specificity of its hiring requirements, then deciding whether that's a problem set they want to own.
Working in AI? Zero G Talent tracks the openings: see every open kapa.ai role, browse AI jobs, the companies hiring, and the people building the field.