As of August 2026, LinkedIn listed roughly 250 open Spoon roles in the United States, with new postings added daily. Zero G Talent's board, which ingests listings directly from the source, shows six current positions spanning design, manufacturing, program management, and facilities, with one role added in the past seven days. They include a Senior Product Design Engineer in San Jose at $230,000–$275,000 (Zero G Talent's board data shows), a Process Engineering Program Manager at $160,000–$230,000, an Office Admin/Manager in Fremont at $39–$45 hourly, plus a Junior Product Design Engineer, an Equipment & Facilities Specialist, and an Equipment Technician in manufacturing. Across the three salaried roles the board tracks, the band runs $97,000–$266,000 with a $230,000 median.
| Role | Location | Band |
|---|---|---|
| Senior Product Design Engineer | San Jose | $230k–$275k |
| Process Engineering Program Manager | San Jose | $160k–$230k |
| Office Admin/Manager | Fremont | $39–$45/hr |
Spoon is actively recruiting across aerospace engineering, satellite systems, launch operations, and mission support using a skill-based AI interview platform that replaces résumé screening with live technical demonstration. The sector spread is deliberate: each discipline is one where a missed spec or a communication lapse can cost millions.
The volume and variety of roles also signal something about Spoon's customers. A hiring surge this broad across launch, satellite, and mission-support functions implies multiple frontier-tech firms are leaning on the same platform simultaneously. The next section examines how that platform evaluates candidates — and why the screen has become the gate that matters.
Inside the AI Interview
Spoon's interview engine runs 30 to 60 minutes of voice conversation per candidate — not a chatbot exchange, not a multiple-choice quiz. Employers upload the questions they would ask in a first round. The AI asks them, then follows up based on each answer, probing deeper the way a skilled recruiter would. The pipeline moves through five stages: Post, Match, Interview, Screen, Connect. A job goes live, the system matches candidates on skills, salary expectations, work type and location, then invites the top matches to an AI-run interview. After the interview, verified screening tests — General Ability and English, timed and auto-scored — layer onto the interview score. The result is a candidate card that leads with skills, experience, interview score and salary fit. Names, photos and contact details stay stripped until the employer chooses to connect.
The scoring logic is deterministic. Verified tests produce fixed scores; the interview itself is scored against the employer's uploaded rubric. Every ranking factor — skills match, interview performance, verified test results, salary alignment — is explainable. That defensibility matters in frontier-tech hiring where a bad call on a propulsion engineer or an ML researcher can cost months. Spoon's credit model makes the economics transparent: AI candidate match costs 25 credits, an interview invite 8 credits, the AI interview itself 50 credits, and connecting with a shortlisted candidate 40 credits. Plans start at $99 per month (Solo) with credits that roll over; annual billing includes two free months. Posting jobs and browsing the anonymized talent pool costs nothing.
For small companies without a recruiting team, the platform runs the entire funnel — sourcing, interviewing, shortlisting — so founders and hiring managers only engage with vetted candidates. For large organizations, it absorbs high-volume first-round interviews, freeing internal recruiters to spend time only on the strongest shortlists. In both cases the anonymization is default, not optional: bias has nothing to grab onto because identity is withheld until the employer actively decides to connect. The system executes a repeatable evaluation pipeline that outputs ranked, scored candidates. Employers still make the hire — they just do it from a shortlist built on demonstrated ability rather than resume keywords.
What the Screen Demands
The platform's stated philosophy is skills-first: candidates build one profile tied to the role they perform best, then sit a 30-to-60-minute AI interview that evaluates "skills, experience and interview, the things that actually predict whether you'll be great at the job." There is no resume upload step; the profile and the spoken interview carry the full weight. Spoon's marketing makes the bargain explicit: "Speak your answers; we clean the transcript" and "Judged on your skills, not your name or photo."
The evaluation architecture draws on structured-interview science: meta-analytic research shows scored interviews reach predictive validity of .51 versus .20 for unstructured conversations. Spoon's AI applies a role-specific scorecard built from weighted competencies. For the engineering and product-design roles currently open, the competency weights typically cluster around technical capability, problem-solving, communication, and collaboration signals — consistent with frameworks recommended by hiring-operations research (e.g., U.S.
Technical capability carries the heaviest weight in standard technical scorecards. But "technical capability" is never a single bucket. The scorecard breaks it into the concrete sub-skills the role actually demands: system-design reasoning, coding fluency, debugging discipline, security thinking, cloud-infrastructure depth, data-modeling rigor, or platform-reliability judgment. A Senior Product Design Engineer interview probes mechanical-design validation and manufacturing-process awareness; a Process Engineering Program Manager screen weights cross-functional program execution and stakeholder translation higher. The AI evaluates each sub-skill against behavioral anchors — specific, observable descriptions of what each performance level looks like — so that a candidate's answer is measured against the same rubric every time.
Problem-solving stands as its own category. The distinction matters: a candidate can write clean code yet freeze when requirements shift mid-project. The AI presents ambiguous, multi-constraint scenarios and scores the structure of the reasoning, not the final answer. Research on behavioral interviewing identifies resilience and adaptability as qualities employers probe, and Spoon's rubric maps to those dimensions.
Communication receives a dedicated weight in technical scorecards and a larger share in early-screen scorecards. The AI evaluates clarity, concision, and audience awareness, specifically whether the candidate can explain a complex trade-off to a non-technical stakeholder without jargon inflation. Organizations using skills-focused assessment data are more likely to make a successful hire, and communication clarity consistently ranks among the top skills measured.
Collaboration and ownership signals round out the model. The behavioral anchors target conflict resolution, cross-functional collaboration, ownership/accountability, and growth mindset. The AI listens for framing, credit allocation, and descriptions of feedback received and acted upon. Similar-to-me bias (the tendency to favor candidates who mirror the interviewer's background) is structurally suppressed because the rubric, not a human, assigns the score.
Every rating requires evidence notes: what the candidate said, built, explained, or solved. The EEOC requires employers to retain all hiring records, including interview notes and scoring tools, for a minimum of one year after the decision (two years for federal contractors), and documented evidence for each rating is the only defensible record.
Pass/fail minimums gate the process before weighted scoring even begins. Minimum qualifications and schedule availability are binary filters. Motivation and role fit, often a significant share of early-screen weight, are assessed through the candidate's demonstrated knowledge of the product trajectory and their articulated reason for choosing this specific problem space. Generic answers score at the bottom of the behavioral anchor scale.
The practical implication for candidates is unambiguous: prepare demonstrable artifacts, not rehearsed narratives. Have a debugging walkthrough ready. Have a system-design decision you owned, with the trade-offs you weighed and the metric you moved. Have a conflict you resolved and the feedback loop you closed. The AI does not reward polish; it rewards evidence.
Candidates Adapt: Proof Over Polish
The old playbook of keyword-stuffing a résumé, optimizing a LinkedIn headline, and hoping a recruiter spots the right brand names is losing relevance inside Spoon's pipeline. The platform's architecture forces a different behavior: candidates build a single profile and undergo the same AI interview process. The interview length, up to an hour, means surface-level fluency won't survive follow-ups. The platform's "explorative follow-ups" function like a persistent senior engineer: they keep pulling the thread until the candidate's actual depth is exposed.
This mirrors what frontier-tech teams have long practiced behind closed doors. PostHog describes its ideal interview as "a conversation between two engineers solving a problem together." Meta's technical screen is a 45-minute conversation with an engineer assessing technical skills. Spoon has productized that dynamic: the AI interviewer doesn't tire, doesn't rush, and doesn't privilege pedigree. The result is a preparation curve that rewards hands-on proof, such as open-source contributions with meaningful commit histories, reproducible project demos, and written design docs, over narrative polish.
The broader hiring market is moving the same direction. Jobs for the Future publishes skills-based hiring toolkits with job descriptions, assessments, and screening guidance for in-demand occupations. HireVue sells AI-powered video interviewing and skill validation to enterprises. Spoon's differentiation is a full-stack replacement of the resume screen, not a supplement. For a candidate, that means the portfolio is the application. A Senior Product Design Engineer role at $230,000–$275,000 will not be won by a degree line; it will be won by a candidate who can walk the AI through a mechanical-constraint trade-off study they led, then defend the material selection under cost pressure.
Candidates are adapting. They treat the profile as a living technical artifact, updating project links, embedding architecture diagrams, and quantifying impact in engineering terms (e.g., "reduced p99 latency 40% by rewriting the scheduler") rather than business-speak. They practice spoken technical communication: explaining a distributed-system failure mode aloud, without slides, in real time. They study the evaluation rubric implicitly; Spoon's promise of "bias-free by design" and "companies evaluate you on your skills, experience and interview" signals that every minute of the interview maps to a scored competency. There is no small talk buffer.
The tension is real: a platform that removes human bias also removes human charity. A rambling but brilliant engineer who would charm a hiring manager may flunk an AI that scores structural clarity. Candidates who recognize this are investing in structured communication, such as frameworks like "context, constraint, approach, result," not as interview theater but as a survival skill. When the gatekeeper is a model trained to detect competence signals, the only reliable hack is to be competent and make it legible.
Why Employers Buy In
Spoon's pitch to employers centers on a single claim: the platform delivers ranked, skills-first shortlists without the bias, time sink, and fee structure of traditional recruiting. The company's own materials frame the value proposition in three layers: objectivity, speed, and cost predictability, each tied to a specific mechanic of the AI interview loop.
Objectivity comes from the interview itself. Every candidate for a given role answers the same voice-based session, with the AI generating follow-up questions based on the candidate's responses. The transcript is cleaned and scored against the role's defined skill rubric. Spoon's site states the result plainly: "Companies get objective, skills-first shortlists and decide on merit, with bias out of the equation." No names, photos, or pedigree signals reach the hiring team until after the shortlist is produced. For frontier-tech firms where talent pools are narrow and referral networks tend to replicate existing demographics, that structural blind spot removal is the primary draw.
Speed follows from delegation. The AI conducts the full first-round interview, including scheduling, proctoring, and probing, without a human recruiter on the calendar. Spoon's marketing emphasizes "AI runs 30–60 minute interviews for you," positioning the platform as a force multiplier for lean hiring teams. A solo founder or a small engineering lead can post a role, let the system interview dozens of applicants overnight, and wake up to a ranked list. The research materials repeat the phrase "AI-matched, skills-first shortlists" as the core deliverable: a decision-ready slate, not a pile of résumés to screen.
Cost predictability is built on the credit model detailed earlier. "Pay in credits, with strict cost guardrails" and "Credits, not headhunters" appear across Spoon's copy, contrasting the platform against contingency recruiters who charge 15–25% of first-year salary. For that role at that salary, a traditional headhunter fee would run $34,500–$68,750 per hire. Spoon's credit pricing, free to start then pay-per-interview, caps exposure before a single candidate is contacted.
The workflow closes the loop at the handoff. "When a company wants to take you forward, Spoon connects you for the next round," the platform steps aside after the shortlist, leaving final assessment and offer to the employer. That boundary matters: Spoon does not claim to replace the hiring decision, only to make the input to that decision cleaner and faster.
No independent employer testimonials appear in the research corpus. The evidence for adoption rests on Spoon's own stated mechanics and the live board activity showing multiple concurrent openings across hardware, software, and operations functions. For frontier-tech firms weighing the platform, the signal is structural: a repeatable, auditable first round that costs a known amount and surfaces candidates who can demonstrate the required skills in a live technical conversation.
The Shift Toward Skill
Recruitment has moved through three distinct phases. The 1990s brought online applications and digital résumés, digital recruiting 1.0. The 2000s centralized job aggregation, 2.0. Today, extensive AI integration defines 3.0, and the adoption curve is steep. A 2019 industry survey found 88 percent of organizations globally had already experimented with AI in recruitment; 41 percent used chatbots for candidate engagement, 44 percent for candidate identification via social media and public data, and 43 percent for training recommendations. By 2022, 86 percent of employers reported using technology-driven interviews, with automated video interviews a preferred method.
The early track record was troubled. In 2015, Google's job recommendation system displayed high-income postings more frequently to men than women. In 2017, Amazon discontinued an AI evaluation tool that systematically scored women's résumés lower because the model had learned gender biases from historical hiring data. In 2019, Facebook's ad delivery system skewed job ads by gender and race. These failures trace to five root causes documented in the literature: biased training data, vague label definitions, feature selection that encodes prejudice, proxy variables that reveal protected attributes, and intentional masking by prejudicial developers. Job-ad text itself compounds the problem: gendered keywords and cultural stereotypes embedded in word embeddings deter applicants and shrink diversity before a single résumé arrives. A 2018 study of Indeed, Monster, and CareerBuilder found significant group unfairness in 12 of 35 job titles studied.
Public trust lagged. A 2019 poll showed 88 percent of Americans skeptical of AI-driven recruitment; as of 2023, 71 percent opposed AI making final hiring decisions. The skepticism was earned. GPT-3.5-turbo (May 2023) reproduced the pro-White callback gap documented in field experiments; White-coded names received callbacks 2.12 percentage points more often than Black-coded names, a magnitude exceeding the 1.6-point within-employer gap measured at 108 Fortune-500 firms. But the trajectory shifted fast. Beginning with Claude 3 Haiku in March 2024, every subsequent model either showed no statistically significant racial preference or, where a difference appeared, favored Black applicants. By 2026, OpenAI's lineage moved from +2.12 pp (pro-White, 2023) to −0.61 pp (pro-Black, 2024) to null. The authors of that analysis conclude the direction of algorithmic hiring bias is not a fixed property of language models but varies systematically with model vintage, provider, and scale, consistent with successive generations of post-training alignment shifting behavior from reproducing pretraining patterns toward neutrality.
Neither the original pro-White bias nor its pro-Black reversal is desirable from a fairness standpoint. A hiring screener that systematically favors one group over another, in either direction, fails the basic requirement of equal treatment regardless of race or gender.
Regulation followed. New York City's law, effective 2023, prohibits automated decision-making tools in candidate screening without prior bias audits and transparency disclosures. The Biden Administration launched the National AI Research Resource Task Force; the Algorithmic Fairness Act (2020) and Algorithmic Accountability Act (2022) advanced at federal level; in 2021 alone, 17 states introduced AI regulations. Technical countermeasures matured in parallel: Textio applies AI to suggest neutral, inclusive language in job ads; researchers distinguish ad-delivery skew from qualification differences; others tackle recency bias in retrieval; generative adversarial networks strip gender bias from word vectors; job recommendation is modeled as a resource-allocation problem with cross-group consistency constraints. The research taxonomy now separates pre-processing, in-processing, and post-processing mitigation strategies against individual, group, procedural, and regression fairness metrics.
Frontier industries feel the pressure acutely. The U.S. aerospace and defense market, valued at $463 billion, is growing faster than it can staff itself. Biotechnology roles are evolving faster than hiring strategies adapt; skills assumed sufficient twelve months ago are already being reassessed. Robotics and automation demand engineers, technicians, and AI specialists at a clip that outpaces traditional pipelines. The World Economic Forum's Jobs of Tomorrow report maps how AI, robotics, energy, and network technologies are reshaping seven major job families employing 80 percent of the world's workers. In defense hiring for 2026, the leading trends are skills-first hiring, accelerated clearance processing, and community-based talent pipeline development, all movements that align with evaluating demonstrated capability over pedigree.
Spoon's skill-based AI interview platform sits inside this convergence. Its screen replaces résumé proxies with hands-on technical demonstration, directly addressing the proxy-variable problem that lets bias slip through even when protected attributes are removed. The platform's merit-first shortlists reflect the same logic driving NYC's audit mandate and the industry's shift toward skills-first hiring: measure what the candidate can do, not where they studied or who they know. The wider shift isn't theoretical; it's regulatory, technical, and economic, and it's moving in one direction.
Two hundred fifty roles sit open today. Each one will be filled by someone who can walk an AI through a debugging session, a trade-off study, a failure mode, and make it legible. The gatekeeper doesn't charm. It measures.
Working in frontier tech? Zero G Talent tracks the openings: see every open Spoon role, browse frontier tech jobs, the companies hiring, and the people building the field.