The Team and the Mission
The researchers who built the neural attention mechanism, ChatGPT, DeepMind's GNoME materials engine, and OpenAI's Operator agent have reunited. Their new company, Periodic Labs, operates on a single premise: frontier AI has exhausted internet-scale training data (roughly ten trillion text tokens) and now needs verifiable, high-signal experimental data to advance. Physics provides the ideal reward function: fast iteration loops, simulators for large classes of systems, and nature itself as the reinforcement-learning environment. This guide covers who Periodic Labs recruits, how compensation works, what the interview process tests, where the work happens, and which traits predict success.
The initial target is high-temperature superconductivity, where the ambient-pressure record sits at 135 Kelvin. Beating that number would signal the system works. But the architecture is designed to generalize across materials domains. An early industry deployment already helps a semiconductor manufacturer solve heat-dissipation problems.
The hiring plan mirrors three interlocking pillars the founders call "fractal" in depth: large-language-model and reinforcement-learning infrastructure, physics-based simulation, and autonomous experimental operations. Each pillar contains multiple sub-teams. On the LLM side, the company seeks researchers and engineers who have innovated in mid-training, RL, and distributed training infrastructure. The simulation pillar splits between theoretical physicists who can formulate problems and software engineers who can scale simulation codes. The experimental pillar, described as the most operationally complex, spans solid-state chemistry, solid-state physics, laboratory automation, and facilities engineering. "For each of these we try to get basically the best people who have innovated in these sub-pillars," a founder said in a 2025 technical discussion.
The scientific advisory board signals the depth of domain expertise the company considers necessary: Carolyn Bertozzi and Zhi-Xun Shen (Stanford Chemistry and Physics), Mercouri Kanatzidis (Northwestern Chemistry), Steve Kivelson (Stanford Physics), and Chris Wolverton (Northwestern Materials Science and Engineering).
The company's own description, "AI scientists and the autonomous laboratories for them to operate," means every hire must function at the intersection of their specialty and at least one other pillar. The founders have said the difficulty of the problem is "partly the team" — assembling a group where these translations happen daily rather than through handoffs. That integration requirement shapes every role description and the compensation structure built around it.
Compensation at the Top of the Market
Job board data shows a salary band running from $100,000 to $350,000 across 21 salaried roles, and the figures put the median at $275,000. Every posted role in the current cohort sits at or above $200,000 base, and the upper bound clusters at $350,000 for engineering and research-track positions.
| Role | Location | Base Salary Range (USD/year) |
|---|---|---|
| Research Engineer — Midtraining | Menlo Park, CA | $250,000 – $350,000 |
| ML Systems Engineer | Menlo Park, CA | $250,000 – $350,000 |
| Founding HR Leader | Menlo Park, CA | $250,000 – $350,000 |
| Product Engineer | Menlo Park, CA | $250,000 – $350,000 |
| Research Scientist, Condensed Matter Theory | United States (remote-eligible) | $225,000 – $325,000 |
| Mechanical Engineer | Menlo Park, CA | $200,000 – $300,000 |
Research Engineers and ML Systems Engineers, roles that sit at the intersection of large-model training and autonomous lab infrastructure, command the same top band as the Founding HR Leader and Product Engineer. The Condensed Matter Theory position, while slightly lower at the ceiling, still starts at $225,000. Mechanical Engineering, the only role with a $200,000 floor, remains above typical hardware-engineering benchmarks for early-stage companies.
Equity detail is not published in the board listings. The investor roster — a16z, Felicis, DST, NVentures, Accel, plus individual backers Jeff Bezos, Elad Gil, Eric Schmidt, and Jeff Dean — implies a standard high-growth venture structure. Candidates should expect the equity component to be meaningful relative to base, given the capital intensity of autonomous-lab build-out and the company's stated goal of creating "AI scientists" that can design materials from superconductors to semiconductor thermal interfaces.
The founding team co-created ChatGPT, DeepMind's GNoME, OpenAI's Operator, the neural attention mechanism, and MatterGen. They are backed to scale labs that generate gigabytes of experimental data per run — data that exists nowhere else and feeds directly into model training. Roles are priced for people who can operate inside that loop.
Geography matters less than function. The Condensed Matter Theory role lists "United States" rather than Menlo Park, and its band overlaps the on-site engineering ranges. That signals remote flexibility for deep-theory talent, but the premium remains tied to the same output standard: verifiable progress in a domain where nature is the reinforcement-learning environment.
For candidates comparing offers, the $100,000 spread between floor and ceiling within a single role compresses negotiation to demonstrated capability at the interview stage rather than title or years of experience.
The Interview Process: What Gets You Through
Periodic Labs does not publish a step‑by‑step interview playbook. A 2026 conversation with leadership frames the search as "decomposed the world into bits and atoms" — mid‑training and pre‑training AI roles on the bits side; control engineering, systems engineering, and product engineering on the atoms side.
The company describes its culture as "physicists and chemists working really closely with some of the top AI researchers in the world, working closely with some of the best engineers in the world." The same video notes that "some of the scientists who joined us are among the best in the world and been absolutely incredible working with them." That is the calibration.
The research does not disclose the exact number of interview rounds, panel composition, or take‑home assignments. What it does make clear is that the process selects for the same interdisciplinary velocity the product demands.
The Workplace: Menlo Park and What It Implies
Periodic Labs lists its open roles in Menlo Park, California — a detail that appears consistently across recent postings for Research Engineer, ML Systems Engineer, Product Engineer, Mechanical Engineer, and Founding HR Leader. Beyond the job listings themselves, public records do not disclose the company's exact address, lab layout, or facility specifications. Periodic Labs has not published a virtual tour, floor plan, or facility blog post that would let candidates visualize the workspace before they interview.
The Menlo Park context matters. In July 2022, Helios Real Estate Partners paid $16 million for a vacant 35,000-square-foot warehouse at 4055 Bohannon Drive with the explicit intention of converting it into a biology lab building on a speculative basis, a bet on the Bay Area's life-science real estate market, which "ranks at or near the top nationally in most key metrics," per The Real Deal. The city had recently rezoned the West Bohannon Park area from general industrial to office district, clearing the way for the conversion. Cushman & Wakefield's Paul, who brokered the deal, expected the project to finish in the second half of 2023. The asking rent was projected in the high $7-per-square-foot range, roughly $2 above Menlo Park's then-average of $5.47. That project illustrates the type of infrastructure being built nearby: shell space retrofitted for wet-lab use, with the ventilation, power, and containment systems that biochemical work demands. Whether Periodic Labs occupies a space like that, a traditional office-lab hybrid, or a different site entirely is not documented in the available sources.
What the research does establish is the kind of work the facilities must support. The company's own site describes a strategy built on autonomous labs that "provide huge amounts of high-quality data (each experiment can produce GBs of data!) that exists nowhere else," starting in the physical sciences because "technological progress is limited by our ability to design the physical world." That mission implies a need for integrated wet-lab stations, robotic sample handling, compute clusters adjacent to the bench, and the ability to iterate hardware (machine-shop access or at least rapid-prototyping capability) without leaving the building. The job postings reinforce the mix: a Mechanical Engineer role suggests hardware design and fabrication needs; an ML Systems Engineer role points to on-premise or low-latency GPU infrastructure; a Research Scientist in Condensed Matter Theory indicates computational work that still sits close to experimental validation.
Berkeley Lab's history offers a parallel for what "integrated" looks like at scale. Since 1931, accelerators have driven its mission, and the Advanced Light Source, launched in 1993, generates intense X-ray beams for experiments across materials science, chemistry, and biology. A major upgrade (ALS-U) aims to increase low-energy X-ray brightness a hundredfold and focus beams to a few billionths of a meter. The BELLA petawatt laser program pursues compact, high-energy acceleration. Those facilities combine beamlines, end-stations, sample-prep labs, and data-analysis halls on one campus. Periodic Labs operates at a different scale and with a commercial timeline, but the architectural logic is similar: co-locate generation, measurement, and compute so the feedback loop stays tight.
Until Periodic Labs publishes more, the only grounded answer is: the work happens in Menlo Park, in a region actively building the lab stock to support it, and the specifics are worth a direct conversation.
Who Thrives Here
Periodic Labs sits at an intersection where the team that co-created ChatGPT, DeepMind's GNoME, the neural attention mechanism, and MatterGen is now pouring that expertise into wet-lab synthesis. The company's own site frames the philosophy bluntly: "Intelligence is necessary, but not sufficient. New knowledge is created when ideas are found to be consistent with reality."
A Technology Review profile of the AI-materials field notes that "startups entering the space are looking to combine computational and experimental expertise in one organization" and that solid-state synthesis, Periodic's starting domain, is "far more difficult to automate than the liquid-handling activities that are commonplace in making drugs." The company's first San Francisco lab is "just beginning to set up... starting with manual synthesis guided by AI predictions; its robotic high-throughput lab will come soon."
Investor Susan Schofer of SOSV put the commercial bar in concrete terms: she wants to see startups "finding something new, that's different, and know how they are going to iterate from there" and a business model that "captures the value of new materials." Periodic's founding team, backed by a top-tier syndicate, has the runway to take that long view, but the board-data postings signal they're hiring for throughput now.
The company's manifesto highlights that autonomous labs "generate valuable negative results which are seldom published." The industry context sharpens the profile. Materials discovery has delivered only a few memorable commercial breakthroughs, lithium-ion batteries among them, in 40 years. The 2023 DeepMind "millions of new materials" claim was later criticized by MIT economists who had "no confidence in the provenance, reliability or validity of the data." Periodic's founders lived that hype cycle from inside DeepMind and OpenAI. They're explicit that the physical world is the RL environment — verifiable, noisy, and slow.
Cross-functional fluency is non-negotiable. The semiconductor manufacturer engagement, "training custom agents for their engineers and researchers to make sense of their experimental data in order to iterate faster," requires translating between fab engineers and ML researchers.
Working in frontier tech? Zero G Talent tracks the openings: see every open Periodic Labs role, browse frontier tech jobs, the companies hiring, and the people building the field.