Job description

ABOUT MARIANA MINERALS Mariana Minerals is a software-first, vertically integrated minerals company on a mission to supply the critical minerals powering modern energy, AI, and defense technologies. We’re reimagining the minerals supply chain by combining deep industry expertise with advanced software, automation, and data-driven decision-making. THE ROLE Mariana Minerals is a software-first, vertically integrated minerals company supplying the minerals critical to modern energy, AI, and defense technologies. Our data doesn't live in a vacuum — it comes off sensors, assays, lab benches, and plant logs, and it's what the business uses to decide what to run, what to buy, what to fix, and what to build next. We're hiring a Senior Data Scientist to be the person who makes those numbers trustworthy and legible. You'll own the statistics: how we measure recovery and throughput, how we know a process change actually worked, how much of a swing is signal and how much is assay noise, and what the reporting layer says when an operator or an executive opens it. This is not an ML modeling role and not a data engineering role. Our MLEs build the models and our data engineers move the data. You bring statistical rigor and clear reporting to a company generating more data than it currently knows how to interpret, and you make sure the conclusions people draw from it are the right ones. WHAT YOU'LL DO - Design and analyze plant trials and experiments — DOE, sample sizing, control selection, and the analysis that says whether a process change did what it was supposed to do and at what confidence. - Own the definitions of the metrics the business runs on: recovery, grade, throughput, yield, uptime, unit cost. Decide what each one means, make the definition consistent across teams, and defend it when someone wants to compute it differently. - Build and own the reporting and analytics layer — the recurring reporting, the dashboards, and the self-serve tooling that lets operators, engineers, and business leads answer their own questions. - Quantify uncertainty honestly: sampling error, assay variability, instrument drift, measurement system analysis. - Apply statistical process control and capability analysis to plant operations — know when a process has actually shifted versus when it's a normal excursion. - Run the deep-dive analyses that don't have a home: why last month's recovery dropped, what's driving cost variance, which of these three suppliers is actually better, whether this correlation is real. - Do forecasting and estimation for production, cost, and capacity planning — applied statistics that informs commitments, not research models. - Partner with the Technical Product Manager for Data & Analytics Platform on what the reporting and analytics stack needs next, and be a demanding internal customer of the data platform when the data isn't fit for purpose. - Raise the analytical bar across the company: review other people's analyses, catch the flawed comparison before it reaches a decision meeting, and teach the teams around you enough statistics to stop making the same mistake twice. - Write findings up so they're actually used — a clear memo a non-statistician can act on, not a notebook that needs you in the room. HOW YOU'LL OPERATE - Rigor with a deadline: You know the difference between the analysis worth another week and the one that's good enough to decide on today, and you say which is which. - Skeptical by default: You ask how the data was collected before you analyze it, and you're the person who notices the sensor was miscalibrated for three weeks. - Structure from ambiguity: Turn a vague business question into a well-posed statistical one, get the data you need, and own the answer end to end. - Explicit about confidence: Communicate uncertainty in a way that helps people decide, rather than hedging so heavily the analysis becomes useless. - Ecosystem fluency: Understand where your data comes from, what the models and simulators downstream do with it, and where the reporting layer sits relative to the platform underneath it. WHAT WE'RE LOOKING FOR Must have - 3–6+ years in data science, statistics, analytics, or a quantitative research role where you owned analyses that drove real decisions - Deep applied statistics: experimental design, hypothesis testing, regression, uncertainty quantification — and the judgment to know which tool fits the question and which assumption you just violated - Strong SQL and Python (pandas, statsmodels, scipy) — enough to get your own data, run your own analysis, and produce your own reporting without waiting on someone else - Track record building reporting and dashboards people actually use, with the product sense to know what belongs on a dashboard versus in a memo - Ability to work with messy, real-world measurement data — missing values, inconsistent sampling, instrument error — and be clear about what it can and can't support - Exceptional written communication; much of this job is making a technical finding land with operators, engineers, and executives - Comfort being the statistical authority in the room Nice to have - Background in mining, metallurgy, chemicals, energy, manufacturing, or other heavy industry — especially metallurgical accounting, mass balance reconciliation, or sampling theory - Experience with statistical process control, or measurement system analysis - Experience with BI and analytics tooling (Hex, Looker, or embedded analytics) and an opinion about how to deploy them - Degree in statistics, chemical engineering, chemistry, operations research, economics, or a related quantitative field - Experience building an analytics function at a company where one didn't exist yet - Working fluency with LLM-assisted analysis and where it genuinely speeds up analytical work versus where it quietly introduces errors WHY THIS ROLE Most senior data scientist roles are one of two things: a modeling role wearing an analytics title, or a dashboard-request queue. This is neither. You own how a physical, capital-intensive business measures itself, at the moment those measurements move from spreadsheets and tribal knowledge into something supported and shared. At Mariana, you don't need to validate product-market fit, because we are the market: if you find something real in the data, the people who can act on it are down the hall. OUR CULTURE IS BUILT ON FOUR PRINCIPLES: Everyone Gets Home Safe. We never put speed or cost ahead of people. Extreme Ownership. We take full responsibility for outcomes, relentlessly driving toward solutions. Engineer Out Requirements, then Automate. We simplify, optimize, and then automate for scale. Share Your Legos. We collaborate openly, share knowledge, and empower each other to build bigger, better solutions. Join us as we build the future of responsible mineral sourcing and supply!

Requirements

Must-haves

Design and analyze plant trials and experiments — DOE, sample sizing, control selection, and the analysis that says whether a process change did what it was supposed to do and at what confidence.
Build and own the reporting and analytics layer — the recurring reporting, the dashboards, and the self-serve tooling that lets operators, engineers, and business leads answer their own questions.
Quantify uncertainty honestly: sampling error, assay variability, instrument drift, measurement system analysis.
Apply statistical process control and capability analysis to plant operations — know when a process has actually shifted versus when it's a normal excursion.
Run the deep-dive analyses that don't have a home: why last month's recovery dropped, what's driving cost variance, which of these three suppliers is actually better, whether this correlation is real.
Do forecasting and estimation for production, cost, and capacity planning — applied statistics that informs commitments, not research models.
Write findings up so they're actually used — a clear memo a non-statistician can act on, not a notebook that needs you in the room.
Rigor with a deadline: You know the difference between the analysis worth another week and the one that's good enough to decide on today, and you say which is which.
Skeptical by default: You ask how the data was collected before you analyze it, and you're the person who notices the sensor was miscalibrated for three weeks.
Structure from ambiguity: Turn a vague business question into a well-posed statistical one, get the data you need, and own the answer end to end.
Explicit about confidence: Communicate uncertainty in a way that helps people decide, rather than hedging so heavily the analysis becomes useless.
Ecosystem fluency: Understand where your data comes from, what the models and simulators downstream do with it, and where the reporting layer sits relative to the platform underneath it.
3–6+ years in data science, statistics, analytics, or a quantitative research role where you owned analyses that drove real decisions
Deep applied statistics: experimental design, hypothesis testing, regression, uncertainty quantification — and the judgment to know which tool fits the question and which assumption you just violated
Strong SQL and Python (pandas, statsmodels, scipy) — enough to get your own data, run your own analysis, and produce your own reporting without waiting on someone else
Track record building reporting and dashboards people actually use, with the product sense to know what belongs on a dashboard versus in a memo
Ability to work with messy, real-world measurement data — missing values, inconsistent sampling, instrument error — and be clear about what it can and can't support
Exceptional written communication; much of this job is making a technical finding land with operators, engineers, and executives

Nice-to-haves

Background in mining, metallurgy, chemicals, energy, manufacturing, or other heavy industry — especially metallurgical accounting, mass balance reconciliation, or sampling theory
Experience with statistical process control, or measurement system analysis
Experience with BI and analytics tooling (Hex, Looker, or embedded analytics) and an opinion about how to deploy them
Degree in statistics, chemical engineering, chemistry, operations research, economics, or a related quantitative field
Experience building an analytics function at a company where one didn't exist yet
Working fluency with LLM-assisted analysis and where it genuinely speeds up analytical work versus where it quietly introduces errors

Application prep

Use the role’s language and application details while tailoring your materials.

Keywords to mirror

aimlsqlpythonllm

Platform: ashby
Effort: quick