What Is a Member of Technical Staff (MTS)? Salary, Roles, and Why AI Labs Are Fighting to Retain Them

The Member of Technical Staff role is one of the most loaded titles in AI right now, and the salary range behind it is wider than most people expect.

Earlier in my career, when I was working at Meta, MTS wasn't a title that came up often. A lot has changed since then. Many of the smartest engineers I know now hold MTS titles across frontier labs, AI companies like OpenAI and Anthropic, and research-driven startups. I've watched friends earn these roles, seen candidates target them, and had countless conversations about what the title actually means in practice. As a tech startup founder myself, I've also spent a lot of time thinking about how exceptional technical talent should be hired, evaluated, and empowered, and we've hired engineers into MTS-style roles ourselves.

At many of the most influential AI companies in 2026, MTS is one of the most important technical titles you can hold, and the absence of a rank is often intentional. Understanding why tells you a lot about how modern AI labs are structured, how they think about engineering talent, and why so many ambitious candidates are suddenly targeting these roles.

Where did the MTS title come from, and why did labs revive it?

MTS started at Bell Labs in the 20th century, the place that produced the transistor and the laser. The idea was to let people move between research and engineering without a title getting in the way.

Anthropic, OpenAI, and others brought it back for the same reason. Anthropic uses a single "Member of Technical Staff" title with no research-versus-engineering split. The flatness is deliberate. With no level attached, people can work across pre-training, post-training, product, and infra without renegotiating their title every time.

There’s also a practical advantage in the middle of today’s AI talent wars. When every major lab is aggressively recruiting from every other lab, generic titles reveal less information about an employee’s exact level, scope, and responsibilities. A title like “Senior Staff Research Engineer, Pretraining” immediately tells competitors who you are and what you do. “Member of Technical Staff” tells them much less. While internal levels still exist, the external title creates a layer of ambiguity that makes it harder for rivals to map organizational charts and target specific talent.

The clearest sign this isn't a junior badge came in May 2026, when Andrej Karpathy announced he was joining Anthropic as an IC on the pre-training team, building a group that uses Claude to speed up pre-training research. Peter Bailis left the CTO seat at Workday, an $8 billion company with 18,000 employees, after less than a year to do RL engineering at Anthropic. CTOs from major companies made the same move. When people who could hold any title pick the flat one, the title means something.

How much does a Member of Technical Staff actually make?

This is where it gets wild, and where you have to read the fine print.

At a frontier lab, MTS comp is enormous. Anthropic base salaries run roughly $300K to $425K depending on level, plus substantial equity. Levels.fyi puts median total comp around $545K, with a reported range from about $198K to $759K. Treat those as directional, not precise, since the public numbers vary and some are internally inconsistent.

Now the catch. The same title pays completely differently depending on where it sits. Median US AI engineer total comp sits around $245K, with the top quartile clearing $350K. So "Member of Technical Staff" tells you almost nothing about pay on its own. The company matters far more than the title.

If you're aiming at frontier labs specifically, the people most likely to land those rooms tend to know someone inside. Simplify Network surfaces your 1st and 2nd-degree connections at Anthropic, OpenAI, and other labs, and lets you request warm intros from employees on the inside. When credentials matter less than the right introduction, your network is your edge.

What does the MTS job feel like day to day?

Careers pages make MTS sound like deep, contemplative research. The reality is faster.

One engineer I know joined a fast-growing AI startup and was told early on that they needed to ramp faster, despite still being in the process of setting up their environment and learning the stack. Within weeks they had shipped a substantial refactor and taken ownership of critical infrastructure work, but changing team priorities and a later reorganization reshaped the situation entirely.

That's the texture of the role. You're expected to ship within days, and the internal mobility that the flat title enables is real but moves fast enough to give you whiplash. If you want a defined ladder and a clear promotion path, know that progression inside MTS is genuinely undefined right now. Even people inside these companies admit they still have to figure out what "next" looks like.

How does MTS hiring actually work?

Here's the part early-career readers get wrong, and it's the part I care about most, because it's the kind of hidden rule I had to learn the hard way. The MTS funnel does not reward credentials the way you'd expect.

Anthropic says on its careers page that a PhD and prior ML experience are not required. Roughly half their technical staff have PhDs, which means half don't. Their hiring page is blunt about what to do instead: "If you have done interesting independent research, written an insightful blog post, or made substantial contributions to open-source software, put that at the TOP of your resume."

So put public work first. When I was first applying to internships as a freshman, I didn't even know what an ATS was, and my resume buried anything interesting I'd built three lines deep. The labs are telling you to do the opposite. That means a few specific things you can do this month:

Implement a paper in PyTorch and document it well. Not MNIST, not Titanic. Pick a real architecture, write a clean README, add an architecture diagram, and explain the parts you got stuck on. The documentation matters as much as the code.

Deploy something rather than stopping at a notebook. The most common portfolio failure is a project that ends in a Jupyter notebook with no deployment link. Build a RAG system with a vector database and a working UI, then put it somewhere people can click on it.

Write about the problems you actually solved. There's a story of two engineers, same company, same title, same four years. One made Staff at $340K, the other landed on a PIP. The difference wasn't talent. The promoted one wrote publicly about the real problems she worked through, and her future manager had already read two of her pieces before the promotion cycle. We've seen the same pattern over and over. Public writing isn't a vanity project. It's how the right people find you before you apply.

The interviews differ too. Anthropic uses Google Colab or Replit with live screen sharing and a 90-minute CodeSignal assessment that has you build something like an in-memory database or a banking system. It's not LeetCode, and AI safety and ethics come up throughout. OpenAI runs a different loop: a 4-to-6 hour virtual onsite with live coding, system design, and ML theory like KL divergence and classifier accuracy bounds. People describe needing to be "a coding machine." The whole OpenAI process averages about 30 days.

One honest note: most named MTS examples are senior people. The clearest entry path for someone without an ML background is the OpenAI Residency, a 6-month program paying $18,333 a month, roughly $220K annualized, built for career-changers from physics, math, neuroscience, or software engineering. No PhD required, and residents do convert to full-time.

If you're aiming at this, the practical stack to show is Python, PyTorch, and some MLOps, with NLP and LLM work front and center, since that's the single most-screened skill in current postings. Build the public proof first, then track which labs and residencies are open and apply the day the window opens.

MTS roles open fast and windows close faster, especially residencies like OpenAI's. Our Job Tracker keeps you organized across multiple frontier labs and alerts you the moment new postings hit, so you're applying on day one, not day five. The students who break into hard funnels aren't usually the ones with the best credentials. They're the ones who understood the timing.

Frequently Asked Questions

Is Member of Technical Staff an entry-level title?

Not really. Most named MTS hires are senior engineers or former executives, so treating it as a new-grad title sets you up for disappointment. The realistic on-ramp for someone without an ML background is the OpenAI Residency, a six-month program that converts strong residents to full-time. Build deployed, public projects before you aim straight at an MTS posting.

How is MTS different from a regular software engineer role?

A standard SWE title signals a level and a ladder. MTS strips both out on purpose, so the same title can cover infra, pre-training, or product. The trade-off is ambiguity: you gain mobility across teams but lose a clear promotion path. You also tend to ship faster, sometimes within your first week, with less structure around you.

Do I need a PhD to become a Member of Technical Staff?

No. Anthropic states a PhD and prior ML experience are not required, and roughly half their technical staff don't have one. What replaces the credential is demonstrated public work: a documented paper implementation, an open-source contribution, or a deployed project with a real UI. The bar is high, but it's a portfolio bar, not a diploma bar.

What should I put on my resume for an MTS application?

Lead with public, verifiable work instead of coursework. Anthropic literally says to put independent research, a sharp blog post, or open-source contributions at the very top. Add links that someone can click: a deployed RAG app, a clean GitHub README, an architecture diagram. Keep tutorial projects like MNIST off the page entirely, since they signal nothing.

How long does the frontier lab interview process take?

OpenAI's full process averages around 30 days and includes a four-to-six hour virtual onsite with live coding, system design, and ML theory. Anthropic leans on a 90-minute CodeSignal assessment plus live screen-sharing rounds with a heavy AI safety focus. Plan for a few weeks of preparation per lab, and track each timeline so you don't miss a follow-up.