What Numen Builds
Numen wants to make cancer a survivable disease by catching it before symptoms show up. The company bets the bottleneck is no longer biology, but access.
The San Francisco startup, founded in 2021 by Thomas Carroll and Luca Springer, builds rapid at-home urine tests that return results in roughly 30 minutes. The platform fuses synthetic biology with machine learning and computer vision: engineered biochemical assays flag cancer biomarkers, and ML pipelines read the resulting signals to classify what's in the sample. The company's own site describes the team as "top experts in bioscience, engineering, ML, and design, obsessed with solving the hardest problem in human health." That blend, wet lab plus production AI, defines what Numen hires for.
Numen went through Y Combinator's Summer 2023 batch under its original name, Cleancard, and is listed on YC's public company page with the founders' bios and the mission line "Engineering a future without cancer deaths." The current scope is narrower than the slogan suggests. Numen's first screens target prostate, bladder, and ovarian cancers, three tumor types where urine-based biomarker detection is already credible, with a turnaround the company describes as ten times faster and cheaper than lab-based tests, and ten times more sensitive than standard lateral flow assays. During its YC batch, Numen reported successfully targeting ten new cancer biomarkers. The longer-term pitch is a multiplexed, at-home platform that tracks biomarkers continuously, not a one-off test.
The founder pairing explains why an 18-person diagnostics startup hires AI engineers at all. Carroll is a clinical cancer researcher with a specialty in bioinformatics and biomarker methods development. Springer holds an MSc in Computer Science and an MSc in Global Governance from Oxford, studied as a Rhodes Scholar, previously analyzed and helped grow healthcare companies at Stockdale Street, and is a Schmidt Futures ISF Fellow. Wet-lab oncology on one side, computational science and healthcare systems thinking on the other; that is the company in miniature. For candidates, the practical consequence is that an AI role at Numen isn't a generic ML position with a healthcare logo stapled on; the bar is whether you can build production systems against biological data alongside people who think in biomarkers and clinical workflows.
The market case the company makes in public materials is straightforward: more than 18 million Americans currently have cancer, with projections of 26 million by 2040, and even high-adoption tests like PSA screening reach only about one in three eligible people. Lab diagnostics are too slow and too expensive to catch pre-symptomatic disease, and current at-home rapid tests aren't accurate enough to trust for cancer. Closing that gap (fast, cheap, sensitive enough to act on) is what the team is hiring to solve.
The Three Open Positions
Three roles make up Numen's current hiring slate, and each one sits at the intersection of applied science and engineering. According to Numen's Work at a Startup page, the company is hiring for three jobs; the specific roles, drawn from YC's job listings for Numen, are:
| Role | Salary range | Equity | Experience |
|---|---|---|---|
| Computational Protein Design Lead | $140K – $220K | 0.10% – 0.25% | 1+ years |
| Visual Designer | $90K – $140K | 0.05% – 0.10% | 1+ years |
| Research Associate II | $75K – $110K | 0.05% | 1+ years |
All three positions are based in San Francisco. The published descriptions sit alongside the founders' profile, the company's funding history, and the Cleancard launch post, the original YC pitch under Numen's earlier name that described a rapid-test method reading cancer biomarkers from urine in 30 minutes with those same speed, cost, and sensitivity advantages. The launch post frames the platform as combining synthetic biology, machine learning, and computer vision into a multiplexable, machine-readable format for detecting multiple cancer biomarkers from a single sample, with prostate, bladder, and ovarian cancers as the first screens.
The YC listing describes the team combining backgrounds in oncology research, machine learning, and healthcare systems to support translational diagnostic development, a hybrid profile that defines the hiring bar. The jobs page on YC does not enumerate every required skill line-by-line in the way the broader public discussion of AI-biotech hiring does; what the public materials support is the company's stated blend of oncology and ML expertise, not a word-for-word job description of each role's screening criteria.
Inside Numen's Screening Process
Public information on Numen's interview process is thin. Glassdoor's Numen interview page lists seven interview reviews and seven interview questions from candidates; regional Glassdoor mirrors for 2024 show four-to-five reviews. The company has not published a formal breakdown of its stages.
What the Glassdoor candidates describe is a recruiter-led conversation up front, followed by a technical round, and then a panel or on-site loop. Anonymous reviewers say the recruiter screen runs short, on the order of 15 to 20 minutes, and reads less like a coding check and more like a filtering call. The technical round, according to the same reviews, includes a take-home or live coding component that goes beyond standard problem solving. Reviewers describe being pushed on topics relevant to medical data, including class imbalance, calibration, and distribution shift between training cohorts and hospital deployment, and asked to walk through a published paper of their own or defend the methodology of a past project.
The through-line in candidate write-ups is consistent: the screen rewards applicants who can marry production-grade machine-learning engineering with grounded oncology knowledge. The available reviews do not quantify pass rates by stage; what they describe is a process that filters for hybrid expertise rather than testing for it in isolation.
What Candidates Say Gets You Past the Screen
The signal from past Numen applicants, drawn from the Glassdoor interview threads, points in one direction: candidates who describe building complete ML pipelines on real clinical data, from raw medical images or omics inputs through to a deployed, evaluated model, repeatedly report advancing past the first round. The reviews describe the screen as a credibility filter for people who can already speak the language of oncology, not a generic coding gate.
The pattern that recurs in candidate write-ups is production evidence rather than theoretical depth. Applicants who cleared the initial screen describe walking through a project where they owned the full lifecycle, including data ingestion, label curation, model training, validation against a clinically meaningful endpoint, and deployment into something a clinician or researcher could actually use. The reviews don't quantify pass rates by stage, but the qualitative through-line is consistent.
Several reviewers point to publications or formal research output as a tiebreaker at the screen stage. The reviews don't say publications are required, but candidates who list them report that recruiters spent more time on the technical conversation. Domain knowledge shows up in subtler ways too: candidates describe being quizzed on cancer-specific concepts and report that interviewers dug deeper into those answers than into framework trivia.
The honest caveat: Numen-specific candidate data is thin, and the company is small enough that individual anecdotes can swing the narrative. What the research supports is the pattern, not a statistic. Applicants who combine end-to-end ML production experience with demonstrable oncology fluency describe advancing; applicants who show up with only one half of that pair describe stalling at the screen.
Why This Screen, Why Now
Numen's insistence on hybrid talent, engineers who can ship production ML and reason about tumor biology, maps onto the same fault line running through the broader AI-biotech labor market, where the scarce commodity is no longer raw modeling skill or wet-lab experience alone, but the bilingual operator who can sit between the two.
The supply pressure behind that filter is intense. AI hiring has accelerated into a winner-takes-most market, with TechCrunch's 2025 tally recording 55 US AI startups raising rounds of $100 million or more across the year. Per Hired in AI's analysis of more than 10,000 postings at OpenAI, Anthropic, Google, and Meta, Senior ML Engineer compensation climbed 15% year-over-year to total packages above $500,000 at top firms, and 45% of AI roles now offer remote options, up from 32% the prior year. One a16z general partner told TechCrunch directly: "Right now, recruiting in Silicon Valley is the hardest thing, because you're competing with Anthropic and OpenAI that have raised billions and billions of dollars for the best talent." For a YC-stage startup pitching a cancer-killing mission against that gravity, the screen has to do real work.
That pressure is layered on top of a biotech sector that has been shedding jobs. Drug Discovery Trends' running map of 2024 cuts had surpassed 20,000 across biotech and pharma by late May of that year, with oncology hit harder than most therapeutic areas. Bristol Myers Squibb trimmed 2,200 roles, Biogen announced 1,000, Amylyx cut roughly 70% of its workforce after the Relyvrio Phase III failure, and Aerovate Therapeutics laid off 39 people (78% of staff) following a Phase IIb flop. The cut-and-hire pattern it documents is striking: biotech startups that were thriving pulled in nearly $3 billion in Q1 2024 funding even as incumbents retrenched. Numen sits firmly on the hiring side of that split, but it is hiring into the same labor pool those laid-off bench scientists and clinical-data engineers are now swimming through.
The hybrid candidate Numen rewards is also what the wider industry says it wants. Forbes' 2024 AI-biotech trends piece flagged clinical validation as the differentiator, urging investors to back companies whose models have been tested against real-world evidence, work that demands operators fluent in both pipelines and pathology. Daily Pulse Magazine's biotech recruitment coverage noted the same shift from "strong scientific backgrounds" toward "a diverse set of skills," with recruiters actively sourcing people who can move between data engineering and bench science. Hired in AI's takeaway for job seekers was blunt: "Focus on practical LLM experience and production deployment skills to stand out in today's market." Practical production work, not theoretical mastery, is the currency, which is the bar Numen's interview loop is built to test.
Remote and cross-border hiring patterns reinforce the point. a16z's Borderless Founder network reflects a market where "the concept of headquarters is something that is changing a lot; it's just more fluid," per partner Gabriel Vasquez's interview with TechCrunch. Smaller AI-biotech teams recruit from Boston, the Bay Area, and international hubs simultaneously, which widens the candidate pool but also widens the screen, because hiring managers now sort through applications from people with very different domain exposures.
One tension worth flagging: the research describes AI hiring as a candidate's market, while biotech oncology has been a candidate's casualty list. Numen's screen demanding proof of end-to-end ML on medical data is, in that light, less about gatekeeping and more about making sure the engineers it does hire can survive a clinical-trial failure cycle that has already ended careers at larger employers.
Working in frontier tech? Zero G Talent tracks the openings: see every open ASML role, browse frontier tech jobs, openings at Stripe, and the people building the field.