Scale AI in 2026: The Founder Left, the Pentagon Stayed, and 180 Roles Are Open

Scale AI is having the most consequential year of its existence. In early 2026, founder Alexandr Wang — who started the company at 19 and built it into a $29 billion AI data infrastructure powerhouse — left to run Meta's new Superintelligence Labs. Weeks later, the Pentagon expanded Scale's defense contract from $100 million to $500 million. And somewhere in between, the company laid off 14% of its workforce and cut 500 contractors.

These aren't independent events. They're the three acts of a company in radical transition: from a founder-led startup that sold data labeling to autonomous vehicle companies, to a Meta-backed, government-entangled AI infrastructure player with no founder at the helm. For the roughly 1,200 employees who remain — and the engineers considering one of 178 open roles — the question isn't whether Scale is interesting. It's whether this particular moment of institutional upheaval is one you want to walk into.

We pulled data from Scale's company profile, employee reviews, compensation reports, and public filings to give you an honest assessment. Here's what we found.

Scale AI at a Glance

These numbers tell a complicated story. A 3.5 Glassdoor rating is below the median for companies in our Culture Directory. The 2.7 work-life balance score is one of the lowest we track. Only 56% of employees recommend the company to a friend. And yet: $29 billion valuation, $500 million in government contracts, and a hiring pipeline that's still running at 178 open positions. Scale is a company where the opportunity is enormous and the experience of actually working there divides opinion sharply.

The Three Earthquakes of 2026

You can't understand what it's like to work at Scale right now without understanding the three events that reshaped the company in rapid succession.

1. Alexandr Wang leaves for Meta

In early 2026, Meta completed a roughly $14.3 billion investment for a 49% stake in Scale AI. As part of the deal, founder and CEO Alexandr Wang departed to become Meta's Chief AI Officer, taking over Meta Superintelligence Labs — a new division absorbing Meta's AGI Foundations group. Scale remains an independent entity, but its founder, its cultural North Star, and the person who coined the company's internal mantra ("Why Not Faster?") now works at Meta.

For employees, this was seismic. Wang wasn't just a CEO — he was the youngest self-made billionaire in America and the gravitational center of Scale's identity. He hired directly, set the pace, and embodied the intensity that defined the culture. Multiple employee reviews describe the post-Wang era as "uncertain" and "disorienting," while others see it as an opportunity for the company to professionalize and build more sustainable management structures.

2. The 14% layoff

Scale cut 14% of its workforce and ended contracts with 500 global contractors. The interim CEO's memo framed it as a restructuring: the data labeling business had "scaled too quickly," creating bureaucracy and confusion about the team's mission. A spokesperson clarified that the cuts were part of a pivot toward enterprise and government contracts, with plans to rehire hundreds of people for those efforts.

The layoffs hit the annotation and operations teams hardest. Engineering was less affected but not immune. For a company that had already been through workforce fluctuations — Scale's contractor workforce has always been large and volatile — this round felt different because it came alongside the founder's departure. The combination created a moment of genuine organizational anxiety.

3. The $500M Pentagon contract

In May 2026, the Pentagon's Chief Digital and Artificial Intelligence Office (CDAO) expanded Scale's defense contract from $100 million to $500 million — a five-fold increase. The deal covers computer vision, generative AI decision support, and data operations for military planning and operations, all running through Scale's Donovan platform, which deploys AI agents on networks classified up to Top Secret.

This contract doesn't just change Scale's revenue mix. It changes the type of company Scale is becoming. Government AI work means security clearances, compliance overhead, slower iteration cycles, and a fundamentally different engineering culture than the move-fast startup ethos Wang built. Scale now has a 13-person Public Sector team and a separate Public Sector Engineering org with 13 more — and both are hiring aggressively.

What the Culture Actually Feels Like

Scale's culture was always intense. Wang's "Why Not Faster?" wasn't aspirational — it was operational. Engineers were expected to own large surface areas, ship without heavy management scaffolding, and operate at a pace that felt more like a seed-stage startup than a company with 1,200 people. That intensity attracted a certain type of engineer: someone who wanted real ownership, wasn't fazed by ambiguity, and could thrive in an environment where the roadmap shifted quarterly.

Employee reviews describe this as both the best and worst thing about Scale. The people who love it talk about the sheer scope of what you can touch — RLHF data pipelines for OpenAI, evaluation infrastructure for frontier AI models, computer vision systems for military applications. The people who struggle talk about the lack of process, inconsistent management quality, and a culture that can feel more competitive than collaborative.

The in-office culture is significant. Scale is headquartered on Market Street in San Francisco with a strong in-person expectation. AI Platform engineers are in-office 4–5 days a week, with some flexibility for focused remote days. This isn't a remote-friendly company. Of the 178 open roles, 40 list "San Francisco, CA; New York, NY" as the location, 29 are SF-only, and 12 are in Washington, DC (government work). The international footprint includes London (9 roles) and Mexico City (8 roles).

Glassdoor Ratings Breakdown

Scale's 3.5 overall Glassdoor rating, based on approximately 500 employee reviews, places it in the bottom third of the 118 companies in our directory. For context, that's the same overall score as Palantir — another company known for government work and intense culture — and well below peers like Anthropic (4.5) or Databricks (4.2).

The 2.7 work-life balance score is the most telling number. Among the companies we track, only a handful score lower. Multiple reviews describe 16-hour days during major client pushes, weekend work as routine rather than exceptional, and an expectation that engineers are "always on" during delivery cycles. This isn't a company that's accidentally intense — it's a company that selected for intensity and is now dealing with the consequences as it tries to scale beyond startup mode.

The 2.9 culture & values score is equally significant. It suggests that the internal experience hasn't kept pace with the external brand. When employees give a company below 3.0 on culture, it typically reflects not just unhappiness with specific policies but a disconnect between what the company says it values and how work actually gets done day to day.

What Employees Actually Say

What employees love

What could be better

The pattern in the reviews is clear: Scale attracts talented engineers with enormous scope, pays them well, and then grinds them hard. The people who thrive are those who genuinely want to be in the trenches of AI infrastructure — building the data pipelines that train frontier models, creating evaluation frameworks that determine which AI systems are safe to deploy. The people who burn out are those who expected the intensity to be temporary rather than structural.

Compensation & Total Comp

Scale pays competitively, particularly at senior levels. Compensation is one of the strongest arguments for joining, especially given the $29 billion valuation and the equity upside potential.

Employee-reported total compensation for software engineers ranges from approximately $226K at the 25th percentile to $699K at the 90th percentile. A senior engineer (L5) typically receives around $425K total comp with a $225K base salary, with equity forming a significant portion of the package. For context, that puts Scale roughly in line with Databricks and above most non-frontier AI companies, though below the top tier of Anthropic ($300K–$490K) and OpenAI ($350K–$550K).

The equity component is the wild card. With Meta holding 49%, Scale's path to liquidity is unusual. It's not a traditional IPO candidate in the near term, and the governance structure — independent company with a near-majority investor who also employs the founder — creates questions about long-term equity value. Engineers who joined before the Meta deal may see their equity differently than those joining now. The $29 billion valuation means new hires get less equity-per-dollar than early employees did at lower valuations.

Engineering at Scale: What You'd Actually Build

Scale's engineering org is organized around its major product lines, and the work is genuinely varied. Here's where the 178 open roles cluster:

• Enterprise Engineering (25 roles) — Building the platform that enterprise customers (OpenAI, Meta, Microsoft) use for data labeling, RLHF, and model evaluation at massive scale
• Research (16 roles) — LLM evaluation methodology, data quality metrics, and the science of what makes training data effective
• Human Frontier Collective (16 roles) — Scale's network of expert annotators and the platforms they work on, focused on high-quality RLHF data
• Public Sector Engineering (13 roles) — Building and deploying Scale Donovan, the platform that runs AI agents on classified government networks
• Gen AI Operations (13 roles) — Operationalizing generative AI pipelines at scale — prompt engineering, quality assurance, and delivery management
• AVCV / Robotics (7 roles) — The original autonomous vehicle computer vision work that Scale was founded on, now a smaller but still active division

The technical challenges are real. Scale processes billions of data annotations. The evaluation infrastructure has to be trustworthy enough that AI labs use it to determine whether their models are safe to deploy. The Donovan platform has to meet military security standards while still being usable. These aren't vanity problems — they're infrastructure problems at the center of the AI industry.

The interview process reflects this intensity. The 2026 loop consists of six rounds: a recruiter screen (30 minutes), a coding screen (1 hour on HackerRank or live), and a four-round onsite covering two coding rounds (medium-to-hard DSA), one system design, one object-oriented design, and behavioral probing. Scale looks for engineers who can operate without heavy management, make decisions under ambiguity, and build systems that handle real-world messiness — not just clean algorithmic solutions. If you're preparing, our Scale AI interview prep guide covers the full process.

The Government Work Question

For many engineers, the $500 million Pentagon contract is the elephant in the room. Scale Donovan deploys AI agents on networks classified up to Top Secret and Sensitive Compartmented Information. The work involves computer vision for military planning, generative AI decision support for operations, and data analysis for national security applications.

This is a meaningful cultural filter. Some engineers are drawn to defense work — it's high-impact, technically challenging, and well-funded. Others have ethical objections to building military AI, particularly in the current geopolitical climate. Scale doesn't hide this work; it's central to their growth strategy. The 178 open roles across Public Sector and Public Sector Engineering make that clear.

If you're considering Scale, you need to reckon with this honestly. Even if you join the Enterprise Engineering team building tools for OpenAI, you're working for a company whose largest single contract is with the Department of Defense. Your work may not directly touch classified systems, but it's part of the same organizational mission. For some people, that's a feature. For others, it's a dealbreaker. There's no wrong answer, but there's definitely a wrong time to figure this out — and that's after you've already joined.

Who Should — and Shouldn't — Join Scale Right Now

Scale is at an inflection point that could go either way. The Meta investment provides financial stability and customer access. The Pentagon contract provides revenue diversification. The remaining 1,200 employees include many strong engineers who chose to stay. But the company's identity is genuinely in flux. It's trying to be a startup, a government contractor, and a Meta subsidiary all at once — and those are three different companies with three different cultures.

For engineers who thrive in ambiguity, the next 12–18 months at Scale could be the most formative experience of their careers. For engineers who need stability, there are companies in our Culture Directory with better Glassdoor scores, better work-life balance, and less existential uncertainty. Both choices are rational. Just make yours with open eyes.