Passage Labs expands to 16 roles, prioritizing systems thinking over coding speed
The Shape of the Push
As of early August, Passage Labs lists 16 open roles on its Ashby board (six in Growth, five in Technology, four in Finance, one in General), signaling a shift in how frontier tech firms evaluate candidates: systems thinking over isolated technical brilliance.
Engradar's daily scrape counted 12 direct-apply roles on August 6, up one from the prior month. Jobscroller tracked 15. The first-party Zero G Talent board shows 16 open roles. The variance across aggregators is real (16, 12, 15), but the direction is not. Passage is growing with intention, not in bursts.
The functional split tells the story. Engradar's taxonomy (engineering (4), finance (2), other (2), operations (1), admin (1), data analytics (1), marketing (1)) maps closely, with Growth decomposing into marketing, operations, and "other." Seniority skews upward: three principal roles, three senior, three unspecified, two director-level, one junior. Median tenure on the board is 28 days. The pipeline is moving.
Geography sharpens the strategy. Ten of 16 Ashby roles are tagged to headquarters. Three sit in India. Two in Nigeria. One in Uzbekistan. Only four are marked remote, roughly one in four, consistent with Engradar. The company isn't defaulting to distributed; it's anchoring at HQ while planting flags in specific talent markets. India and Nigeria appear repeatedly. Uzbekistan shows up once. The pattern suggests deliberate regional bets, not opportunistic remote hiring.
Compensation data is thin. Jobscroller reports a median posted salary of $200,000 across the five of 15 roles that disclose pay, ranging from $150,000 to $200,000. The Director of Finance role on Zero G Talent lists an hourly range of $8–12, a figure that almost certainly reflects a contractor or part-time structure, not a full-time executive package. The discrepancy underscores how little of the picture is public.
The tech stack attached to Technology roles reads like a modern AI infrastructure shop: Python, Scala, TypeScript, React, Django, GCP, GraphQL, LLM Agents. The presence of Scala and GraphQL alongside LLM tooling signals a team building production-grade systems, not prototypes. The latest Zero G addition, a Principal Software Engineer at headquarters, aligns with the seniority skew toward principal and director hires.
Founded in 2021 by Meral Arik and Zac Choi, Passage describes its mission as "building technology that removes barriers blocking talent from opportunity, combining the latest advancements in AI with human judgment." The current wave (heavy on Growth, anchored at HQ, skewed senior) looks like that mission operationalized. The question is what the screening process actually filters for.
Beyond LeetCode and Papers
Passage's open roles (Director of Data, Principal Software Engineer, Director of Finance, Global Growth Lead (India), Senior Accountant/Finance Manager) read like a conventional scaling startup's org chart. But the job descriptions and the product point to a different filter than the standard algorithmic grind.
The Director of Data posting asks for 6–10 years in analytics engineering, data engineering, or advanced analytics, plus a quantitative bachelor's. That baseline is table stakes. What distinguishes the role is the implied scope: the hire will own the data architecture behind a platform that "turns scattered documents, emails, and edge cases into structured, decision-ready applications." The same language appears on Passage's product page, which describes AI agents that "connect your systems and policies to deploy in days." The job isn't to optimize a model in isolation; it's to design data flows that survive messy, real-world inputs, the kind that break clean benchmarks.
That pattern repeats across engineering and product listings. The Principal Software Engineer role sits at the intersection of platform reliability and AI-agent orchestration. The Global Growth Lead roles, listed twice for India, signal a go-to-market motion requiring navigation of regulatory and operational variance across jurisdictions, not just scaling a funnel. The finance hires suggest a company building financial infrastructure for its customers, not just managing its own books.
Passage's product philosophy: "AI should generate signals and structured data, while human interviewers interpret context and make the final hiring decision" mirrors the screening logic the company applies to itself. Sherlock AI frames its outputs explicitly as "signals, not verdicts." Resumly notes that organizations relying entirely on automation "risk bias, poor candidate experience, and missed soft-skill insights." Broader research shows 40% of candidates feel uneasy about AI in hiring and 45% worry about algorithmic bias, while 26% reject offers due to unclear communication about recruitment stages, a dropout driver Parakeet AI identified. Passage's careers portal emphasizes "24/7 multilingual AI counseling and real-time document and eligibility screening"; transparency about which stages are automated versus human-reviewed would directly address that friction.
What the job descriptions collectively screen for is judgment in ambiguous AI contexts: the ability to decide when a model's output is trustworthy, when to escalate to a human, and how to structure systems so that boundary is legible. That's a different muscle than LeetCode hardness or publication count. It's the muscle Passage's product demands of its users, and, by extension, of the people building it.
The Feedback Loop: Human-in-the-Loop as Hiring Metaphor
Passage builds an AI-augmented talent platform. Its own hiring sits inside the same feedback loop the company sells: software that uses machine learning to surface candidates, then hands decisive judgments to people.
The industry's language for this model has hardened into consensus. Creative Alignments calls it "keeping human judgment in the moments that matter most." People Science frames a "Recruiting Continuum" where AI handles high-volume data processing, resume parsing, and funnel diagnostics while humans own contextual decision-making, cultural assessment, ethical oversight, and relationship-building. Tally AI describes recruiters evolving into "strategic advisors who interpret insights, guide hiring managers, and ensure that hiring decisions align with organisational goals." Rent a Recruiter distills it to three imperatives: use AI to remove administrative work, keep humans responsible for hiring decisions, prioritise candidate experience.
The logic is structural, not sentimental. When everyone uses large language models to write resumes and cover letters, and employers use them to write job descriptions, the signal-to-noise ratio collapses. Creative Alignments documents the result: an "AI doom loop" where more automated applications trigger more automated screening, eroding trust on both sides. Recruiters spend more time filtering resumes than evaluating people. Strong candidates disengage. SHRM finds nearly three-quarters of HR professionals believe AI will increase — not replace — the importance of human judgment, particularly in empathy, ethics, and contextual decision-making. The market has already hit the wall of pure automation.
Passage's product roadmap (inferred from the roles it is filling) maps directly to this architecture. A Director of Data and a Principal Software Engineer suggest investment in the ranking, matching, and explainability layers that make AI outputs auditable. The Global Growth Lead roles in India signal a push into markets where proactive sourcing, not inbound volume, determines outcomes; Creative Alignments reports roughly 90 percent of its client hires come from targeted outreach rather than applicant pools. The finance hires indicate a company preparing to scale revenue operations behind a platform that charges for judgment-augmented matching, not raw resume throughput.
The feedback loop works in two directions. The platform's design choices — where to insert a human review gate, how to surface confidence intervals, what explanation a hiring manager sees before they accept or reject a match — become the same criteria Passage applies to its own candidates. Systems thinking that can articulate why a retrieval-augmented generation pipeline should flag low-confidence matches for human review demonstrates the exact judgment the product requires. Fluency in building evaluation frameworks for fairness and drift detection proves capability in the governance layer that People Science and Tally AI both identify as the next competitive frontier: "accountability remains human, transparent, and auditable."
This isn't unique to Passage. McKinsey's 2026 enterprise AI survey shows 88 percent of organisations now use AI in at least one business function, and the highest returns accrue to those pairing automation with strong governance, human oversight, and change management — not to those deploying AI as a standalone efficiency tool. The companies hiring well in this cycle, across AI, defense, and robotics, are the ones treating the human-in-the-loop boundary as a product decision, not a compliance afterthought. Passage's hiring slate reads like a team building exactly that boundary.
Candidate and Recruiter Reactions: Praise, Confusion, Attrition
The shift toward AI-augmented screening has triggered a feedback loop no hiring team fully controls. Candidates now deploy large language models to optimize résumés, tailor cover letters, and rehearse interview answers, prompting recruiters to report a flood of applications that read alike. Fast Company documented the pattern: repeated phrases, parroted job-posting language, and a "too-polished tone" that signals synthetic authorship rather than genuine fit. Harvard Business Review went further, arguing that generative AI has "broken hiring" by undermining the reliability of traditional signals (résumés, writing samples, remote interview performance) that teams once treated as proxies for competence.
On the candidate side, transparency complaints cluster around opacity. Passage's own help center acknowledges the friction: "The reason for your application's rejection is clearly outlined in your account at app.passage.com. We aim to provide detailed explanations to help you understand the decision." That such a statement exists suggests applicants routinely ask why they were cut. Public forums echo the frustration. A 2026 YouTube tutorial on using NotebookLM for interview prep captured the mood: "Interviewers are impressed when they can see that you've done your homework," the creator said, describing how deep research on panelists' backgrounds shifted conversations from interrogation to dialogue. The implication: candidates who master the AI research loop advance; those who don't, stall.
Recruiters observe a different metric: drop-off. Blind-hiring experiments, including one cited by HireTruffle where identity was stripped until the final offer, found that anonymization alone does not fix attrition. "Anonymizing resumes might increase your top-of-funnel diversity, but without structured interviews, bias creeps right back in," Equalture CEO Charlotte Melkert said. The same research showed that language patterns leak gender with 99 percent accuracy even on name-blind résumés — men default to nouns like "equity" and "capital," women to "organized" and "volunteer" — and that models trained on those corpora amplify the signal unless actively retrained. For a platform building AI-augmented matching, that finding is operational: the screen must be structured, not just anonymous.
Tech communities debate whether the new filters favor a specific profile. Reddit threads on blind hiring, spanning 2018 to 2023, reveal a split. One camp argues the approach levels the field for candidates without pedigree signals; another calls it a "band-aid" that delays bias until the live interview, where "the same biases will apply." A 2023 commenter noted that gaps from caregiving (disproportionately borne by women) still surface in anonymized timelines, and that referrals, "sacred" at most firms, recycle homogeneous networks. Candidate Labs, a specialized recruiter, responded by designing around the problem: "We send fewer candidates, with more context, and we get sharper as we calibrate," their site states, emphasizing tight feedback loops and shared Slack channels over volume.
Passage's product thesis (that AI should augment human judgment in talent matching) mirrors the structured-interview consensus emerging from research: standardized scorecards, consistent questions, and AI surfacing evidence rather than making decisions. But the research contains no Passage-specific drop-off rates, candidate NPS scores, or recruiter testimonials. Whether the company's pipeline bends the attrition curve remains an open question the next hiring cycle will answer.
The Frontier Tech Shift Toward Judgment-Centric Hiring
The shift Passage is riding didn't start with them. Across the frontier tech stack — AI labs, cloud platforms, and the defense and robotics contractors downstream — hiring is being rewired around a single premise: routine code generation is now a utility, so the scarce human skill is deciding what to build, what to trust, and what to ship.
The numbers make the turn undeniable. Draup analyzed 2.85 million job descriptions posted between June 2025 and June 2026 and found software engineering, data engineering, and DevOps roles each still topped 40,000 active listings. But more than 60,000 listings across nine technical categories explicitly named AI tools (GitHub Copilot, Cursor, Claude) as baseline expectations. HackerEarth reports aptitude assessments have surged sharply since 2024 while overall assessment volume held steady, a signal that employers are testing for reasoning, not recall. CodeSignal says most U.S. software engineers now use agentic AI coding tools at work, and a large share have already shipped code that was at least partly generated by AI.
The big platforms moved first. Google is piloting an AI-assisted coding interview that lets candidates use Gemini inside a CoderPad environment (three-panel layout, file explorer, editor, chat window) and evaluates "AI fluency, including prompt engineering, output validation, and debugging skills." The round introduces a code comprehension format where candidates read, debug, and optimize real code with Gemini. Candidates still face a classic algorithm interview without AI. Meta rolled out its own AI-enabled interview in October 2025, letting candidates choose among GPT, Claude Sonnet, Claude Haiku, Gemini, and Llama, and scoring on problem solving, code quality, verification, and communication. Meta's format asks candidates to work with a multi-file codebase across bug-fixing, core implementation, and optimization phases. Canva redesigned its questions to be "more complex, ambiguous, and realistic… the kind of challenges that require genuine engineering judgment even with AI assistance." Problems "can't be solved with a single prompt; they require iterative thinking, requirement clarification, and good decision-making."
The driver is visible in production data. Google CEO Sundar Pichai disclosed in April 2026 that 75 percent of all new code at Google is now AI-generated and approved by engineers, up from 50 percent last fall. OpenAI president Greg Brockman told a Sequoia Capital event that AI coding tools went from writing 20 percent of code to 80 percent "over the course of December" alone. When the typing is automated, the review becomes the job. Draup's analysis of over one million software job descriptions confirms it: debugging and judgment during code review remain essential, while writing boilerplate and recalling syntax are increasingly automated. The same report names systems design, data governance, and model evaluation as skills that stay critical across roles.
That compression hits junior talent hardest. Draup notes expectations for early-career hires are rising fastest because the routine tasks that traditionally served as training ground are now the most automated. The firm argues organizations should help junior workers develop design, review, and judgment capabilities within months rather than years, a fundamental shift from task-based to capability-based workforce planning. Morgan Stanley agrees AI coding tools are more likely to change the shape of software work than eliminate it, pushing developers toward strategic responsibilities while CIOs plan to increase software spending in 2026.
But the trend line has counter-currents. Airbnb CEO Brian Chesky told CNBC AI hadn't changed hiring plans "as much as I thought"; efficiency gains in customer service netted out against accelerated business growth. The Federal Reserve's Beige Book cites AI as one factor making employers cautious, alongside higher interest rates and the hangover from 2021–2022 hiring binges. TechRadar, citing an Oliver Wyman global CEO study, found only 17 percent of chief executives expect to increase junior hiring while 43 percent expect to reduce it. Tom's Hardware reported almost half of tech layoffs in Q1 2026 were attributed to AI and automation. Some executives and analysts caution companies may be using AI as a convenient explanation for wider cost-cutting.
The tally remains 16. The screen for those roles — systems thinking over LeetCode speed — is the same filter the company's product applies to every candidate in its pipeline. The frontier firms that master that filter will set the pace. The ones still optimizing for coding puzzles are hiring for a job that no longer exists.
Working in frontier tech? Zero G Talent tracks the openings: see every open Passage Labs role, browse frontier tech jobs, the companies hiring, and the people building the field.