Working at KoBold Metals: Culture, Pace and Who Thrives

How work actually gets done

In July 2026, KoBold's DRC head Benjamin Katabuka reported his team had digitized 170,000 pages of Lubumbashi university records since April — a single data ingestion feat that captures the company's model: remote‑first, archive‑hungry, and operating where the peace agreement between the DRC and Rwanda (signed June 2025, reaffirmed in the Washington Accords of December 2025) has not stopped fighting on the ground. KoBold runs a global mineral exploration business with a headquarters in Berkeley and a distributed workforce across multiple time zones. The organizational center of gravity sits with the founders and a small executive team, while field operations in the Democratic Republic of Congo, Zambia, and other jurisdictions run through local leads like Katabuka.

Two workstreams coexist in tension. The core R&D loop: geologists and machine‑learning engineers curate government‑funded geological surveys, colonial‑era maps, and drill logs (264,000 aerial photographs, 25,000 maps, 160,000 rock samples, and a kilometre of archives in the Belgium collection alone) into a proprietary model that ranks exploration targets. The other stream is physical digitization and logistics: scanning, OCR, QA, and metadata tagging of fragile paper records in facilities the company does not own. The World Bank warned in 2025 that KoBold would "likely have exclusive access to the data for some time" after digitizing the Zambian Geological Survey, a dynamic repeating in the DRC.

Decision‑making authority is not documented publicly. The cap table — Breakthrough Energy Ventures (Gates), Bezos Expeditions, Jack Ma, Michael Bloomberg, among others, implies a board expecting commercial returns on a funding‑round timeline, not a geological one. The company raised $195 million at a ~$1 billion valuation in 2024, then $537 million at nearly $3 billion in 2025‑26. That capital velocity forces hiring at speed. Whether product calls route through a CTO, a VP of Exploration, or a founder‑level technical council is not disclosed.

The shift from pure exploration to actual mining, flagged in 2024 reporting as a strategic pivot, will add permitting, ESG compliance, and operational safety to a stack that today is mostly code and contracts. Public filings do not reveal whether KoBold has adopted a formal framework (OKRs, RACI, sprint cadences) or relies on ad‑hoc alignment. The only documented rhythm is the funding cycle.

Values and operating principles

The name reaches back to German folklore. A kobold is a household spirit — hausgeist, literally the "authority of the house" from kob (chamber, hut) and walt (ruler, power). Medieval Germans carved foot‑tall figures from boxwood and wax, dressed them in bright clothes, and set them in a corner of the room. The spirit performed helpful chores in exchange for offerings but turned mischievous when slighted. Miners later blamed the kobold for a silver‑colored ore that yielded no metal and released arsenic fumes; the troublesome rock took the creature's name, and the element cobalt derives from it. The folklore carries a quiet operating logic: unseen labor, foreknowledge of danger, small practical powers bound to one hearth or shaft — and a relationship that depends on courtesy. The founders chose the name deliberately.

In a long‑form interview, co‑founder Josh Goldman described the founding thesis: the energy transition requires electrifying the global light‑duty vehicle fleet by mid‑century, which demands all current world resources of copper, cobalt, nickel, and lithium plus an additional five trillion dollars' worth of those materials. The easy deposits have been found; humanity has walked the earth noticing unusually colored rocks for thousands of years. The remaining ore bodies are concealed beneath rock cover, invisible to traditional surface methods. Goldman's background, a doctorate in experimental atomic physics and quantum computing, shaped the response: "our thesis is we have to use data much more effectively to increase our probability of making discoveries and develop better methods of finding the next wave of deposits that are more difficult to find."

That thesis translates into three operating principles. First, integration over silos: "it's an entire system for integrating lots of data and applying a whole range of different analytical methods to guide each of the decisions that we make." Geophysics, geochemistry, satellite imagery, historical drill logs, and academic literature feed a shared probabilistic framework rather than sitting in discipline‑specific repositories. Second, quantification over intuition: "a critical component of doing this is integrating many different types of data together and using quantitative statistical methods to estimate the probability of finding ore deposits in any given location." The industry's typical hit rate — roughly zero discoveries per 100 exploration drill holes, is treated as a baseline to beat, not an immutable fact. Third, time‑horizon honesty: Goldman explicitly frames the new‑materials window as 20 to 30 years before recycling becomes the primary feedstock for EV batteries. "We're not going to need so much more new cobalt and nickel and lithium forever," he said. "We're going to need it for the next 20 or 30 years until we electrify the whole vehicle fleet and then the incremental needs will not continue to grow in the same way." That finite horizon shapes capital allocation: the company acquires and explores properties with the intent to develop or partner, not to hold indefinitely.

Public reporting on how these principles manifest day‑to‑day is thin. The company does not publish a values page, and public employee‑review data is too sparse to extract reliable patterns about cultural execution. What exists in the record is the founders' own articulation, repeated across interviews and investor materials: a data‑first exploration model, a climate‑urgency mandate, and a deliberate rejection of the industry's acceptance of low discovery rates. Whether those principles survive contact with scaling remains untested by public accounts.

What the hiring bar selects for

KoBold's hiring pattern merges two talent pipelines that rarely overlap: senior engineers and data scientists from consumer‑tech giants, and geoscientists who have spent careers in the field. The company's own leadership describes the early team as "frontloaded … with data scientists and software engineers from places like Google, Apple and Meta" who then "worked with geoscientists to develop algorithms and technologies to assemble and analyze large datasets" (Berkeleyside, June 2026). That origin story still shapes the bar today.

Zero G Talent's live job board shows the split in practice. KoBold is hiring across engineering, data, legal, and commercial functions at senior compensation levels. Zero G Talent's data shows the board's overall salary band runs $90k–$256k with a $200k median across 23 salaried roles.

Zero G Talent found that range sits above traditional exploration shops but below the top quartile of pure AI labs.

What unifies those disparate titles is a working model the founders have described repeatedly: a "duet of data scientists or of technologists and geologists working together on the same problem of like a hypothesis and then like validation or invalidation" (Goldman, No Priors interview). Candidates who pass the loop demonstrate they can operate inside that duet. For the technologist side, the signal is production‑grade engineering: building sensor hardware (the hyperspectral imaging system was designed and flown on a light aircraft in less than a year), stitching together structured and unstructured geological corpora into a single queryable system, and retraining models daily from field observations. For the geoscientist side, the signal is comfort with uncertainty: the team deliberately goes "to places where the models are highly uncertain" because "the new training data, small amount of additional ground truth can dramatically improve the predictive power of our models" (Goldman, ibid.). Field time is non‑negotiable; a reviewer who only knows Jupyter notebooks will not clear the bar.

The hiring process also filters for a specific risk tolerance. Exploration success rates have collapsed industry‑wide — "eight discoveries as of 30 years ago and today less than one … high‑quality economic deposit" (Goldman, ibid.), and KoBold's portfolio strategy assumes most of its 60+ projects will not become mines. The company targets low‑cost, high‑impact exploration and expects staff to kill projects fast when data says stop. That mindset shows up in the business‑development and legal roles: the VP of Business Development must structure earn‑in agreements like the Libra deal (CAD$33M over six years for a 75% interest) where KoBold funds exploration but the partner operates the field program, and the employment counsel must support a remote‑first workforce spread across Eastern Time, Zambia, and Australia without creating compliance drag.

Finally, the bar selects for people who treat data as the scarce resource, not the metal. "The scarce resource is not lithium or copper metal in the ground. It's actually information … the information about where the ore deposits are located" (Goldman, ibid.). Candidates who have built data pipelines that ingest century‑old hand‑painted linen maps, satellite gravity gradiometry, and daily drill‑core logs, and who can explain how each source shifts a posterior probability, are the ones who get offer letters.

What current and former employees say

Public review data for KoBold Metals is absent from the research provided. The materials supplied contain extensive historical and folkloric background on the kobold myth, including Germanic house spirits, mine sprites, the etymology that gave the metal cobalt its name, and the creature's later appearances in Dungeons & Dragons and fantasy media, but no Glassdoor summaries, Blind threads, LinkedIn testimonials, or attributed quotes from people who have worked at the company. The job‑board data confirms the company is hiring actively in those roles at those levels, but it does not substitute for the lived experience of the people already inside.

Who thrives here and who burns out

No public Glassdoor reviews, attributed employee interviews, internal surveys, or first‑person accounts from current or former KoBold staff describe daily life, management style, or burnout signals. The company's public footprint consists almost entirely of funding announcements, exploration results, geopolitical reporting, and the job postings on this board. That absence is itself a data point — KoBold has operated with unusually low employee‑facing transparency for a venture‑backed startup of its size and vintage.

What we can infer comes from the documented operating model. KoBold is remote‑first across multiple time zones (job postings specify "Remote (Eastern Time Zone)" or "Remote (Global)"), which selects for people who structure their own days without office rhythms. The work sits at the intersection of three high‑pressure domains: machine learning on sparse geological data, field operations in politically complex jurisdictions (Zambia, DRC, Greenland), and capital‑intensive mine development backed by investors who expect step‑change returns. The salary bands, including VP Business Development at $250–350k per Zero G Talent's figures, Engineering Manager, Data Systems at $200–240k, Software Engineer at $150–240k, signal senior‑heavy hiring; the median posted band is $200k. Junior roles are rare. That suggests the organization expects new hires to operate with minimal ramp and limited mentorship infrastructure.

People who tend to thrive in this profile, based on the documented model, share a cluster of traits: comfort with ambiguous, long‑horizon problems where the feedback loop from model to drill result takes months or years; willingness to travel to remote sites (the Zambia copper discovery, the DRC archive digitization, Greenland field seasons) on irregular schedules; and the ability to translate between geologists who think in rock and ML engineers who think in tensors. The "Data Scientist - (Global)" role at $140–260k explicitly spans that translation layer. Candidates who have shipped models into production environments where ground truth is expensive to acquire — not just benchmark datasets, match the hiring signals.

Burnout risk concentrates where those same traits collide with structural friction. Geopolitical reporting documents KoBold negotiating access to colonial‑era archives in Belgium, digitizing Zambian Geological Survey records under World Bank scrutiny, and securing "strategic partnership" status with the Congolese government while M23 rebels seize mineral concessions nearby. Employees on those workstreams handle export controls, data sovereignty disputes, and security protocols that change with each diplomatic shift. The company's pivot from pure exploration to mine development (the Mingomba copper project in Zambia) adds operational complexity, including permitting, ESG compliance, community relations, and logistics, that pure R&D teams haven't owned before. In a remote‑first org, the people who absorb that complexity are often the same ones building the data pipelines and models, with no local operations buffer.

Zero G Talent's board data shows 23 salaried roles posted across a $90k–$256k band (median $200k). That headcount is small for a company valued near $3 billion with $537M raised. Lean teams in high‑stakes domains tend to reward generalists who own problems end‑to‑end and penalize specialists who wait for handoffs. The flip side: generalists without hard boundaries accumulate scope until something breaks. No public employee accounts exist to confirm whether KoBold has instituted guardrails — rotation policies, mandatory disconnect periods, field‑deployment limits, or whether the mission narrative ("critical minerals for the energy transition") functions as an implicit pressure to overcommit.

The kobold folklore that gave the company its name ends with a warning: the spirit stays helpful only as long as the relationship holds. KoBold has completed that archive digitization, flown a hyperspectral imager built in months, and raised half a billion dollars to turn probability maps into mines. The next funding cycle will test whether the unseen labor, the archive hunger, and the courtesy binding field leads in Lubumbashi to engineers in Berkeley can survive the shift from exploration to extraction.

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