Careers at Scale AI: Teams, Pay and How to Get Hired

Who Gets Hired: Where They Land

Scale AI's public hiring record confirms six senior roles priced between $252,000 and $378,000 across four U.S. metros — but no org chart, no team charters, and no role-by-role pay bands below the director level exist in any verified source. The company's careers page collapses all openings into three buckets: engineers, researchers, operators. The university portal adds internships and new-grad roles "depending on job postings and department needs." The interview guide, dated August 2025, lists role categories that trigger distinct tracks: software engineering, machine learning, data science, product, and "etc." — an open-ended tail hinting at unlisted specializations. The verified record answers what titles exist and where they sit geographically. It does not answer which team does what or how the org actually functions.

Board data shows six active salaried postings, all with base ranges between $250,000 and $380,000. The titles are specific: Director of Engineering, Physical AI (San Francisco); Manager, Research Scientist (San Francisco, New York); Tech Lead Manager, MLRE/ML Systems (San Francisco, New York); Tech Lead, ARC Team (San Francisco, St. Louis, New York, Washington DC); Staff Software Engineer, Public Sector (same four metros); Senior Staff Frontier Agents Engineer (San Francisco, New York). Five of six carry Staff, Senior Staff, Tech Lead, or Director titles — a seniority tilt the board's 150-plus-role aggregate confirms but does not explain.

Glassdoor reviews corroborate the variance. Candidates report processes that differ "significantly across roles." Some move smoothly through phone screens, technical loops, and final rounds. Others describe disorganization and recruiter ghosting. A recent YouTube walkthrough of "real interview questions pulled from real hiring loops" targets senior backend, machine learning, and platform engineers working on agent systems — the planner-executor pattern, multi-agent versus single-agent trade-offs, orchestration overhead. The questions "separate the senior engineers from the mid-level ones," implying a hiring bar that rises with system-design depth.

The "ARC Team" and "Physical AI" labels suggest product- or domain-oriented groupings, but no public document maps their charters, reporting lines, or headcount. The "Public Sector" designation on the Staff Software Engineer role hints at a vertical slice, yet the same title appears across four metros without clarification on whether those are separate teams or a distributed one. The Manager, Research Scientist posting sits alongside the Tech Lead Manager, MLRE/ML Systems, both in San Francisco and New York, but the research/engineering split remains opaque. The university portal's promise of "world-class mentors" and "high-impact projects from day one" is a recruiting claim, not a team taxonomy. No sourced document names a hiring manager, a team lead, or a project code name.

The Pay Picture

Posted Ranges for Senior Roles

The board data beats any third-party aggregate because it reflects live listings ingested at the source. All six postings are senior or leadership positions anchored in major metros — San Francisco, New York, St. Louis, or Washington, DC.

These six listings cluster at the top: every band reaches at least $360,000, and three touch $378,000. Floors range from $252,000 to $302,000. The spread shows Scale prices senior IC and management talent at a premium — but also reveals what the data doesn't cover. No entry-level, mid-level, or non-technical bands appear in the verified postings.

Aggregate Board View

Across more than 150 salaried roles, the overall band runs roughly $75,000 to $331,000 with a $257,000 median. That median sits near the floor of the six senior postings above, suggesting the board skews toward experienced hires. The $75,000 floor likely reflects university or early-career programs; the careers site lists a Technical Advisor Specialist part-time internship in San Francisco, but no public posting with that figure is currently captured.

Benefits: What the Company States

Scale's careers page groups benefits into four categories:

• Health & Wellbeing: health coverage including medical, dental, vision, and mental-health services, and PTO policies that "ensure you'll get time off when you need it to relax and recharge." The company notes offerings "may vary by region as we strive to respond to the unique needs of Scaliens around the globe."
• Personal & Career Growth: annual learning-and-development stipend, leadership breakfasts, manager training, speaker series, employee-resource groups.
• Building Scale Community: office guest policies, happy hours, game nights, book clubs, employee-led events.
• Parental Support: leave policies described as "adequate leave policies" to promote healthy home and work life.

Benefits: What Third-Party Aggregation Shows

Teamblind's independently researched compilation (July 2026) lists a deeper menu, though the site cautions data "may vary by role, location, or other factors." Notable line items:

• Insurance & Wellness (11 items): health, vision, dental, HSA, FSA, life, AD&D, occupational accident, disability, mental health, gym membership, employee-assistance program.
• Retirement (2 items): 401(k), pension plan.
• Financial (3 items): performance bonus, employee stock purchase plan, stock options.
• Time Off (10 items): PTO, volunteer time, sick days, paid holidays, maternity/paternity leave, family medical leave, bereavement, unpaid extended leave, military leave, sabbatical.
• Workplace & Lifestyle (10 items): free lunch/snacks, work-from-home, flexible hours, commuter benefits, pet-friendly office, social events, charitable gift matching, travel concierge, employee discounts, mobile-phone discount.
• Development (3 items): tuition assistance, professional development, apprenticeship program.
• Other (2 items): legal assistance, company car.

The two sources align on the big pillars — health, equity, time off, learning; but the third-party list is more granular. Neither breaks benefits down by role, level, or geography beyond Scale's regional disclaimer.

What Remains Undocumented

No public filing, company blog, or aggregated review site provides:

• Role-specific base/bonus/equity splits for individual-contributor tracks: software engineer, ML researcher, operations, recruiting, sales.
• Geographic differentials beyond the four metros in the six postings.
• Signing-bonus norms or refresh-grant schedules.
• Hourly or contract rates for the annotation and evaluation workforce powering Scale's Data Engine.

Review sites mention salaries in passing but not in structured, verifiable bands. If you're targeting a Director, Staff, or Tech Lead role in a major hub, the board gives you a concrete range. For every other level and location, the verified public record is silent — you'll need to ask a recruiter or wait for an offer letter.

Inside the Interview Loop

Scale AI's interview process follows a recognizable structure, but the experience varies enough that candidates should prepare for inconsistency. ProgramHelp's 2025 guide maps a seven-step sequence: application screening, recruiter phone screen, technical screen or online assessment, a role-dependent take-home assignment, onsite or final interview loops, behavioral and leadership rounds, and offer negotiation. The full timeline typically spans three to six weeks, though scheduling and role urgency can stretch or compress it.

Glassdoor reviews paint a split picture. The company's overall rating sits at 3.6 out of 5 across more than 400 reviews, roughly in line with the IT average, but interview-specific feedback suggests the process hasn't been standardized to the point of predictability. Some candidates describe smooth progression through phone screens, technical interviews, and final loops with clear communication. Others report disorganization, chaotic scheduling, and ghosting from recruiters after apparently positive rounds.

The technical evaluation differs sharply by role. For machine-learning and data-science tracks, ProgramHelp documents a one-hour ML round covering model selection, data preprocessing, and practical optimization cases. Software-engineering candidates face a one-hour coding test at medium algorithmic difficulty emphasizing time complexity and clean implementation. A behavioral questionnaire runs roughly 30 minutes and asks for STAR-method examples across past projects, conflict resolution, and career plans. The hiring-manager conversation is an in-depth project discussion that can run long.

Take-home assignments are role-dependent. One example: build a black-box system around an LLM that asynchronously receives requests, splits them into hundreds of segments, calls the LLM synchronously per segment, and returns results via notification. Another: read two CSV files, convert to structured JSON, use a provided LLM API to classify a column, write results back. A third: a multi-file debugging exercise with locked "error-free" functions and three test cases; find and fix logic errors in task-assignment code with course prerequisites.

For senior backend, ML, and platform roles, the YouTube walkthrough from an interviewer describes agentic-system design questions from actual hiring loops. One prompt asks whether to use a single-agent or multi-agent architecture for automated customer-support ticket resolution: reading, classifying, retrieving documentation, generating responses. The expected answer clarifies constraints: volume, latency budget, compliance requirements, then reasons through trade-offs. Single-agent works for linear workflows, minimal branching, lowest latency, simple failure domains. Multi-agent fits when stages need isolation (classification failures shouldn't break generation), different model or tool capabilities per step, independent scaling (classification at 10x generation volume), or parallelization across subtasks.

A second senior question probes the planner-executor pattern: why separate a planner that decomposes high-level goals from executors that carry out subtasks? The key insight is failure isolation: if an executor fails, the planner can retry with a substitute or abort gracefully; if the planner fails, the entire workflow state is lost. Candidates who treat these as purely theoretical exercises tend to cap out; interviewers look for implementation-aware reasoning.

ProgramHelp's preparation advice emphasizes practical signals: submit take-home solutions with clear documentation, unit tests, and detailed comments; be ready to discuss and optimize your own assignment in the follow-up; review ML fundamentals and have concrete optimization cases ready; keep coding-test solutions tidy and complexity-conscious. The recruiter screen remains the first filter; candidates who clarify constraints upfront and ask about volume, latency, and compliance before designing tend to signal the right instincts.

The company's careers page frames the pitch broadly: "Join Scale and power reliable AI systems. We're hiring engineers, researchers, and operators." Internship and new-grad roles open depending on posting and department needs. But nowhere in the sourced material does Scale publish a unified rubric, a public scorecard, or role-specific pass rates. Candidates enter a process structured on paper but variable in practice; prepare for the documented rounds, but expect cadence and communication to depend on the team and recruiter you draw.

Four Cities Anchor the Footprint

Scale AI's verified physical footprint centers on its San Francisco headquarters, confirmed by BuiltIn's company profile as of April 2026. That source lists 523 total employees, a headcount that aligns with a mid-stage company still concentrated in its founding city rather than dispersed across a sprawling campus network. The careers page advertises nearly 350 roles across 26 departments, but location data on those postings is sparse. Most default to "San Francisco" without specifying whether they sit in a single tower, multiple floors, or a distributed floor plan.

Zero G Talent's board fills the gaps with recent postings that name additional cities. The six senior roles confirm Scale maintains a meaningful presence in four metros: San Francisco, New York, St. Louis, and Washington, DC. The St. Louis and DC listings tie to public-sector and defense-adjacent work, consistent with the company's federal contracts.

Beyond those four cities, the research offers no verified office addresses, lease details, or expansion announcements. No sourced photos of floor plans exist. No mentions of lab space for robotics or hardware integration. No detail on whether the San Francisco HQ houses data-annotation operations or is purely an engineering and product hub. The careers site emphasizes mission — "develop reliable AI systems for the world's most important decisions", and product lines (RLHF, 3D point-cloud annotation, LiDAR/RADAR pipelines, the Scale GenerativeAI Platform) without linking any to a specific facility.

The absence of facility detail is not accidental. Scale's two-tier workforce model (full-time engineers earning $184,000 to $292,560 base per BuiltIn, alongside a large contractor pool for data labeling) suggests physical office requirements differ sharply by role. Contractor work is frequently remote or performed in third-party centers Scale does not disclose. Full-time roles in the board data are tagged with city pairs, implying hybrid or multi-site arrangements, but the research contains no policy statement on days-in-office, badge access, or dedicated secure spaces for government programs.

In short, the verifiable record stops at four named metros and a headcount figure. What those offices look like, how they are equipped, and whether Scale operates specialized infrastructure (edge compute clusters, sensor-fusion labs, SCIF-rated rooms) remains undocumented. Candidates should treat the city list as the only grounded location signal and ask directly about workspace specifics during interviews.

Who Thrives Here?

No official Scale AI communications (blog posts, leadership interviews, public competency frameworks) spell out the traits the company says it values. Instead, nearly 250 Blind reviews (3.2 out of 5), more than 400 Glassdoor reviews (3.6 out of 5), and a Jobs by Culture analysis from June 2026 converge on patterns describing who tends to succeed, or at least survive, inside the organization today.

Intelligence and low ego appear together across multiple Blind reviews. "The people remaining are still highly intelligent and low ego. People are actual friends with each other and willing to collaborate," one reviewer wrote. Others called out "generally smart folks" and "smart people to work with." The Jobs by Culture analysis echoed this: "Smart, driven colleagues" and "smart people, high impact work, and lots of growth opportunities." The signal is consistent enough to treat as a de facto hiring filter: Scale's core teams expect a high baseline of technical ability and a collaborative style that does not rely on hierarchy or posturing.

Ownership orientation shows up repeatedly. Reviewers describe "infinite stuff to own," "a lot of work to do, easy to shine," and "high impact work." The Jobs by Culture piece framed the company as suiting "people optimizing for career velocity and proximity to frontier AI who can absorb a demanding pace." Work-life balance scores 2.3 out of 5 on Blind and 2.9 out of 5 on Glassdoor, the lowest sub-score in both datasets. Long hours around client and model-delivery deadlines are the "most recurring theme" in the Glassdoor corpus. Candidates who need predictable schedules or strong cultural guardrails will find the environment hostile; candidates who treat ambiguity and workload spikes as leverage for visibility tend to rate the experience higher.

Seniority preference has shifted. Blind reviewers observed that "the company shifted towards hiring way more senior engineers a year ago. The fruits of this are showing in process and technical improvements." The same thread noted "a ton of new experienced hires at the executive level, as well as lots of strong incoming ICs." This aligns with the board data showing roles such as Director of Engineering, Physical AI ($302k–$378k), Manager, Research Scientist ($290k–$363k), and Senior Staff Frontier Agents Engineer ($288k–$360k), all senior-to-staff level. The organization appears to be selecting for people who can operate with minimal scaffolding, a trait that becomes necessary when "shifting priorities, evolving structure, and uneven manager experiences depending on which team you land on" are described as standard operating conditions.

Tolerance for organizational complexity is another de facto requirement. Scale's model includes a large contractor and operations workforce alongside its core teams, which adds complexity and blurs how teams collaborate. Reviewers flagged that "the BUs don't communicate with each other and hurts the company heavily" and described the company as "a bag of random smaller companies." People who thrive here handle cross-functional friction without waiting for top-down resolution, a trait overlapping with the "low ego" and "ownership" signals above.

Public Sector (PubSec) draws a distinct subset of reviewers who rate the experience more positively: "PubSec is the best part of the company. There are no layoffs." That business unit appears to offer more stability and clearer mission alignment, attracting people who value those attributes over the higher-velocity, higher-churn commercial or Gen AI teams.

The research provides no statement from Scale AI leadership (CEO Jason Droege, the people team, or hiring managers) codifying these traits into a hiring rubric. No "bar raiser" framework. No competency matrix tied to interview scorecards. The 63% recommend rate on Glassdoor and the 3.1 out of 5 culture sub-score (per Jobs by Culture) suggest the employee base itself is split on whether the current environment reflects a coherent set of values or simply the residue of rapid scaling. Until the company publishes its own framework, the traits above remain inferred from survival patterns, not declared intent.

The verified record — six priced roles, four metros, a headcount, a benefits menu — fits on a single page. Everything else candidates need to know lives in the gaps between sources.

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