The $280k Fellow.app Role Requires This Unlisted Skill

Current Openings at Fellow.app

Fellow.app has six salaried roles on its board, with one added in the past seven days, per Zero G Talent's board data. The salary bands stretch from $35,000 to $280,000, median $188,000 — a split that reveals two distinct hiring engines.

Only one listing appeared in the last seven days.

Core Competencies the Market Now Demands

The six visible roles span a wider remit than a pure AI product team. Only the VP of Software and Automation Engineer map cleanly to engineering and product tracks. The brand and China-based quality roles signal simultaneous pushes on go-to-market and hardware. What the board doesn't show, and no first-party Fellow job description details, is the exact competency matrix the hiring team uses. The company's careers page and interview rubrics are not in the provided sources.

The absence forces a read-across from the broader AI talent market. The Washington Post surveyed companies "aiming to fuel their businesses with artificial intelligence" in October 2025 and found four net-new titles: knowledge architect, orchestration engineer, conversation designer, and human–AI collaboration leader. Each bundles a distinct skill cluster now appearing in job specs across the sector.

Knowledge architect roles demand fluency in retrieval-augmented generation pipelines, vector-database selection, and evaluation frameworks that keep hallucination rates measurable. Stanford HAI's 2024–25 graduate fellows illustrate the depth expected: Peter West works on "unlocking abilities in compact models and characterizing challenges that even the largest models continue to face"; Alan Wang builds "deep learning algorithms for medical imaging… with an emphasis on improving interpretability, robustness, and fairness"; Angelina Wang specializes in "machine learning fairness and algorithmic bias." These are not academic curiosities — they are the daily debugging surface for anyone shipping LLM-powered features to production.

Orchestration engineer maps to the automation and infrastructure layer. The Purdue "Advances in Human-AI Collaboration" volume devotes chapters to "sentiment analysis and language models," "fact-checking beyond machine learning and predictive accuracy," and "situation awareness of AI models in relation to human performance." An engineer who cannot instrument latency budgets, design fallback chains for model degradation, or version prompts alongside code will not clear a technical screen at a company shipping real-time meeting intelligence.

Conversation designer sits at the product–design intersection. The same Purdue text covers "chat-based customer service, covering theory-based interventions to enhance public services and examples of intentional human-technology interaction" and "strategies for moving beyond passive writing assistance." Fellow's meeting-assistant product lives in this exact space: summarization, action-item extraction, follow-up drafting. A designer who treats the LLM as a black box rather than a material with controllable failure modes will produce brittle UX.

Human–AI collaboration leader is the most senior cluster. Stanford HAI fellows Tasha Kim ("developing algorithms and computational frameworks that can enhance purposeful, human-centered cooperation with AI agents") and Stefan Stojanov ("building computer vision systems guided by our knowledge about the generalization, adaptability, and efficiency of human perception") exemplify the research-to-product translation this role requires. The Penn State Faculty Upskilling Fellowship, funded at $300,000 across 20 faculty in November 2025, explicitly targets "human-centered designs of all AI systems" and "large language models and AI integration" — language that now appears in principal-level job descriptions.

What it does contain is a convergent signal from three independent fronts: industry job-title creation (Washington Post), academic fellowship selection (Stanford HAI, Penn State), and systems-engineering codification (Purdue). These competencies now constitute the baseline for any AI product team hiring at Fellow's salary bands. Candidates who cannot point to a shipped feature where they owned the RAG pipeline, the orchestration layer, the conversation flow, or the human-in-the-loop evaluation will struggle to explain how they meet the bar the market has set.

Screening and Interview Process

Fellow's push to fill six roles puts its screening pipeline under pressure, but the company's actual interview architecture isn't documented in public sources. What the research shows is how high-growth AI teams tend to structure evaluation when the talent bar is this high, and where the friction points lie.

Glassdoor's multi-country analysis of more than 154,000 paired interview and company reviews found that interview difficulty correlates with later employee satisfaction. On a five-point scale, the sweet spot sits at 4 out of 5 — difficult enough to filter effectively, not so brutal that it signals dysfunction. A 10 percent increase in interview difficulty associates with a 2.6 percent lift in job satisfaction afterward, though the effect varies: strongest in Australia (3.6 percent) and Canada (3.0 percent), weakest in France (1.5 percent) and Germany (2.4 percent). Interviews rated 1 out of 5 tend to produce poor matches; 5 out of 5 often reflects a toxic culture rather than rigor.

For technical roles, the strongest signal comes from practical examination. SpaceX's hiring lead said: "One of the best indicators of a candidate's ability to perform the tasks we're hiring them to do is to give them a practical examination of sorts." Welders weld during the interview. Engineers solve engineering problems. Resumes, the same source noted, "gauge the ability to write a bulleted list of achievements, and that's not always indicative of success — the resume is not going to be sitting in the seat doing the work on Monday." The same logic applies to AI product roles: a candidate who has shipped LLM features, wrestled with prompt evaluation, or built retrieval-augmented generation pipelines can demonstrate that work far more convincingly than they can list it.

The structure of those evaluations varies. SpaceX candidates report three to ten interviews before an offer — phone screen with a recruiter, then the hiring manager, future teammates, and often a panel. Gartner's process runs five stages: recruiter, team manager, key initiative leader, subject-matter expert (a future colleague), and a group interview where candidates prepare a 10-page document on a pre-selected topic. Writing ability, the Gartner source emphasized, "is probably the most important part of the job." Honesty under pressure matters too: "Admitting when you can't answer something is actually a plus. Don't try to wing it."

Ghosting has become its own metric of process health. Glassdoor reviews by UK workers show mentions of employer ghosting tripled since pre-pandemic, up 208 percent since 2019. Candidates who were actively recruited report the highest ghosting rates. SpaceX applicants describe completing multiple rounds, including phone screens, team interviews, and even onsite visits, then hearing nothing. One Falcon Manufacturing Development Engineer candidate logged five steps before silence: online application, recruiter scheduling, recruiter phone screen, team phone screen, then no denial email, no closure.

For Fellow's six open roles, spanning VP of Software, Creative Director, Director of Integrated Brand Marketing, Automation Engineer, and two China-based quality/compliance positions, the screening burden is real. The board's salary bands imply vastly different evaluation depth. A VP of Software loop at the top of that band typically involves architecture reviews, leadership scenarios, and reference checks that take weeks. An Automation Engineer loop leans heavier on coding exercises and system-design walkthroughs. The China-based roles add time-zone coordination and potentially language assessment.

What candidates can control: ask for a timeline in the first conversation. Business Insider's reporting on ghosting recommends nailing down next steps before the first call ends: "At the end of the interview, the hiring manager will normally ask if you've got any questions. That's a good time to ask about the next steps." If the agreed deadline passes, follow up once, professionally. If silence persists, treat it as data and move on. The market for AI meeting-assistant talent is tight enough that a process that respects candidates' time tends to win the ones worth keeping.

Culture Fit and Remote‑Work Expectations

Fellow's six open roles are anchored in two locations: four in San Francisco, two in Shenzhen/Guangzhou, per that board data. The geographic split matters. The board's salary bands range from $35,000–$45,000 for the China-based roles to $240,000–$280,000 for the VP of Software in San Francisco, a spread that reflects both market differences and the company's current hiring footprint.

What it doesn't show, and no public source confirms, is an explicit remote-first policy for these openings. Fellow.ai's own website emphasizes product-level governance: "Built to the standards of regulated industries," "Compliant by default," "Your meeting data stays yours," and detailed controls for SOC 2, HIPAA, GDPR, and CCPA compliance. That messaging is directed at customers in private equity, hedge funds, and investment banking. It describes the product's trust architecture, not the employer's work model.

The broader research offers context on how remote work has reshaped talent flows, particularly in California. PPIC reports that about two-thirds of the nearly three million Californians who telework full-time hold at least a bachelor's degree, and over half of higher-income Californians who left the state during the pandemic reported working from home — a share that remains far above pre-pandemic levels. The same data shows California's net loss of higher-income adults over the past decade at 165,000, less than 2% of the current total, while lower-income departures were far more substantial. For a company hiring in San Francisco at $185,000–$280,000 bands, the candidate pool is increasingly comfortable with distributed work — but also increasingly mobile.

Candid.org, a nonprofit with nearly 40% of its team remote and the rest often hybrid, illustrates how organizations structure asynchronous collaboration: weekly one-on-one check-ins, daily team standups, occasional in-person outings, and a dedicated "Candid University" for professional development. In the nonprofit sector overall, 19% of job listings are remote and 38% hybrid. Those figures are a benchmark, not a proxy for Fellow, but they signal what experienced knowledge workers now expect.

The tension is practical. Fellow builds an AI meeting assistant that records, transcribes, and summarizes conversations across Zoom, Teams, Google Meet, and in-person settings, with zero-day retention options and MNPI redaction for regulated clients. The product exists to make distributed meetings auditable and actionable. Whether the team building it operates under the same async discipline, defaulting to written updates, recorded decisions, and searchable meeting intelligence, is not documented in the available sources. Candidates should treat the absence of a public remote policy as a signal to ask directly: how does the engineering team use its own tool? How are design reviews conducted across time zones? What does "San Francisco" mean for day-to-day presence?

The board data's location specificity suggests these roles may require physical proximity, at least for the San Francisco cluster. But the product's entire value proposition, "one governed AI meeting notetaker, everywhere your team meets," argues for a culture that dogfoods asynchronous workflows. If Fellow hires for product impact and remote-work fluency as the data contends, the interview process should surface evidence of both. The research doesn't confirm that it does. It confirms the openings, the locations, the pay bands, and the product's compliance posture. The rest is a question candidates need to pose.

Market Impact: Demand for AI Meeting‑Assistant Talent

The AI agents market is expanding at a pace that makes Fellow's six-role hiring push look less like an isolated sprint and more like a calibrated response to a structural shift. MarketsandMarkets puts the 2024 market at USD 5.26 billion, the 2025 figure at USD 7.84 billion, and the 2030 forecast at USD 52.62 billion — a compound annual growth rate of 46.3 percent. Within that trajectory, the vertical AI agents segment is projected to lead with a 62.7 percent CAGR through 2030, while the coding and software development agent category sits at 52.4 percent. Meeting assistants occupy a defined slice of this landscape, and adoption signals from the category's largest incumbent suggest the demand curve is steeper than the top-line numbers imply.

Zoom's AI Companion, the most widely deployed meeting-assistant product in market, is now enabled on more than 1.2 million accounts as of October 2024, with nearly 57 percent of Fortune 500 companies running the feature. Active users grew 40 percent quarter-over-quarter in the same period. Those figures translate directly into engineering load: summarization, action-item extraction, cross-platform context retrieval, and multi-language support all require sustained model tuning, prompt infrastructure, and low-latency serving — exactly the work Fellow's open roles target.

That compensation spread reflects a market where specialized AI product experience commands a premium over generalist full-stack backgrounds. The same MarketsandMarkets report notes that professional service firms, such as law, consulting, and IT services, represent the fastest-growing end-user segment for AI agents, a signal that meeting-assistant tooling is moving from internal productivity wedge to billable-workflow infrastructure. North America leads adoption today, but Asia Pacific is accelerating fastest, driven by enterprise digitalization mandates and government-backed AI funds. Fellow's San Francisco concentration for its highest-paid roles aligns with that geographic weighting.

Competitive pressure is visible in the vendor landscape. Google has surged to the leader quadrant in Gartner's 2025 Conversational AI Magic Quadrant, while Kore.ai and Cognigy hold positions — Cognigy is reportedly nearing acquisition by NICE. SoundHound acquired Amelia in August 2024. Microsoft is embedding agents across Dynamics 365 and GitHub Copilot. NVIDIA released AI Blueprints for multi-step agentic applications. OpenAI launched ChatGPT Gov for U.S. agencies. Each move expands the surface area for talent competition: the engineers who can ship reliable, context-aware meeting intelligence are the same cohort building coding agents, support agents, and vertical workflow automators.

Fellow's surge, with six roles across engineering, product, and design, is a leading indicator that the meeting-assistant subcategory has crossed from feature-layer experiment to product-layer commitment. The board's salary bands, the market's 46 percent CAGR, and the incumbent's 1.2 million enabled accounts converge on one conclusion: the talent pool with hands-on LLM-to-product shipping experience in this domain is thinner than the demand curve requires. Candidates who can demonstrate shipped meeting-summary pipelines, prompt-evaluation frameworks, or asynchronous collaboration tooling will price above the median. Those who cannot will compete for the shrinking share of generalist roles.

What Fellow's Hiring Signal Means for AI Talent Trends

Fellow's six-role push, spanning engineering, product, and design with salary bands across that range, mirrors a market inflection that McKinsey's QuantumBlack division has tracked since 2022. Their July 2024 survey of 1,491 organizations across 101 countries found that hiring difficulty for AI-related roles dropped for eight of twelve categories year over year, yet a slight majority still report trouble filling them. The translator category, a proxy for specialized AI-adjacent work, saw "difficult" or "very difficult" ratings fall from just over 70% in 2022 to just under 65% in 2023 to roughly 55% in 2024. Senior partner Lareina Yee attributed the easing to workers proactively upskilling and corporate training investments starting to pay off.

The deeper shift is structural. Deloitte's 2026 Human Capital Trends research, drawn from 100 C-suite leaders, shows that 59% of organizations take a tech-first approach to AI, yet those doing so are 1.6 times more likely to miss return targets than peers pursuing a human-centric model. Competitive advantage has migrated from model access, now replicable, to the human edge: adaptivity, creativity, and judgment amid uncertainty. Fellow's requirement for asynchronous collaboration fluency isn't a perk preference; it's a proxy for the "dynamic orchestration" capability Deloitte identifies as a primary success driver. Organizations that continuously reconfigure capabilities around outcomes outperform financially and turn volatility into opportunity.

Goldman Sachs estimates AI could replace the equivalent of 300 million full-time jobs globally, while the World Economic Forum projects 85 million displaced by 2026. Yet McKinsey's simulation shows AI delivering roughly $13 trillion in additional economic activity by 2030, about 16% higher cumulative GDP, with 70% of companies adopting at least one AI technology. The net effect isn't elimination but transformation: at least 14% of employees globally may need career changes by 2030, and roles involving repetitive tasks will automate while new positions emerge in AI development, data analysis, and cybersecurity. Fellow's product-focused roles, including VP of Software, Automation Engineer, and Creative Director, sit squarely in that creation tier.

Remote-work fluency operates as a leading indicator. Deloitte frames the shift from static plans to dynamic orchestration as one of three tipping points reshaping work, alongside human×machine collaboration and value creation over cost efficiency. Companies that treat discontinuity as momentum, redesigning work, roles, and value in real time, will set the benchmark. Fellow's distributed team, hiring across San Francisco and Shenzhen/Guangzhou simultaneously, embodies this orchestration: capability deployed where talent lives, not where headquarters sits.

The signal for candidates is unambiguous. The half-life of technical skills is compressing; the S-curve of adoption that once played out over decades now bends in years. McKinsey's data shows the translator role evolving as AI language solutions alter both demand and supply — a pattern repeating across every AI-adjacent specialty. Workers who learn, adapt, and apply new skills directly in the flow of work, supported by organizations building always-on adaptability, avoid stalled transformations. Those who don't face the 30% automatable threshold PwC projects for the mid-2030s.

Fellow's surge is a localized reading of a global rewiring. The companies winning the talent war aren't posting the highest salaries, though Fellow's $240,000 ceiling for VP of Software competes, but designing roles where human judgment compounds AI leverage. The next hiring wave will filter for that compounding ability explicitly. The board's six open roles, one added last week, are the current evidence.

Working in frontier tech? Zero G Talent tracks the openings: see every open Fellow.app role, browse frontier tech jobs, the companies hiring, and the people building the field.