Senior Analytical Data Scientist

Department: Intelligence
Location: San Diego, CA (in-person preferred; frequent on-site presence required if not local)
Type: Full Time
Reports to: Senior Principal Machine Learning Engineer

About Satomic

Satomic’s mission is to close the gap from idea to molecule with faster navigation of chemical space. We are building an automated chemistry platform as an “API to Atoms” by integrating synthetic chemistry taste, laboratory robotics, software, and AI to transform small-molecule synthesis and drug discovery.

Satomic has raised a $15M Seed from fantastic investors including Riot Ventures, HOF Capital, and Compound VC, and our team includes alumni from leading large pharmas, small biotechs, and AI labs.

About the Role

Every reactivity model and synthesis protocol Satomic builds rests on labels derived from analytical chemistry signals. This role is critical to Satomic’s mission, as closing the gap from idea to molecule depends on measuring our own experiments correctly, automatically, and at high-throughput.

Satomic runs thousands of synthesis experiments a day, and nearly every one produces a molecule we haven’t made before. The existing analytical chemistry team has built high-throughput physical methods and an instrument stack with multiple detection methods, taking us from well to raw data. This role builds the models and workflows that turn that data into labels: what each well contains and at what concentration. Most products have no authentic standard, the analytes vary widely, and the volume is far beyond what people can review by hand.

The core of this role is owning that pipeline. While simple, rule-based processing handles many injections, others require judgment (co-elutions, unexpected products, instrument drift), and there are too many to check by hand. Once the pipeline runs reliably, you will help lead our work on standard-free quantitation of diverse analytes at high-throughput, an open problem in analytical chemistry.

If you have built pipelines in a modern untargeted or semi-targeted metabolomics lab, this will feel familiar; we have a high-throughput small molecule LC-MS analysis stitched together by a scriptable, open data pipeline. There is one helpful advantage to our need: we know what went into each well, so the space of possible products and side products is enumerable and constrained.

What You Will Own

Automated interpretation directly from raw instrument output. Peak detection, product identification, adduct and fragment reasoning, feature annotation across the whole sample, and conversion estimates delivered to the reactivity ML team with uncertainty attached. You will build agentic pipelines that process whole campaigns end to end, so analysis capacity grows without adding headcount.
The analysis pipeline itself. How raw data moves from instrument to label: what runs automatically, what gets flagged, what the failure modes are, and how things break or improve as throughput grows and the chemistry expands.
Data quality. You will decide which results go to the ML team and which get flagged or held back. You tell instrument faults from unusual chemistry, develop analytical QC policies, and record each processing choice so others can trace how your results were produced, including ensuring instrument and batch context travels with the date, so an instrument or batch level error is caught before it reaches the model.
Over time, predicting analyte-specific responses from chemical descriptors. CAD/UV/MS response factors and retention time for compounds that have never been run.
Feeding those predictions back into acquisition. Shorter methods, focused gradients, internal standard placement, and multiplexed injections while maintaining data quality.
We have built our own tools for spectral visualization, annotation, and semi-automated review. You will decide how those need to evolve and develop, working with dedicated software engineering resources to build it.

What this Role Is (and Is Not)

This role is:

A computational role. Our analytical chemistry team owns the instruments, the methods, and the detectors, and takes us from a well to raw data. You own everything downstream of that: the models, the pipelines, and the labels. The lab is open to you; if you want to run a confirmatory experiment yourself, or need data faster than the instrument queue allows, you can.
A domain-expert role where you set the technical direction for how Satomic reads its own instruments, and work closely with our analytical chemistry lead, because better predictions change which methods are worth running, and new methods change what the models can learn.
A role with real unsolved problems in it. Predicting CAD response from structure has no published answer, and we have the compounds to build the dataset.
A role with a team around it. You will work alongside an existing lab-based analytical chemistry team who have built out our current instrument stack, methods, and analysis pipelines; a cheminformatics scientist focused on chemical space representation and experiment design, a reactivity ML scientist who builds the predictive models your labels feed, and software engineers who build and maintain the data infrastructure, databases, and tools you specify.

This role is NOT:

A bench or method development role. Recent hands-on experience is welcome and will make you better at the job, and day-to-day instrument work stays with others on the analytical team.
A software engineering role. You will not be expected to build production software yourself. You will help prototype, but production software is built and owned by our engineers. Your job is to specify what should be built clearly enough that engineers and coding agents build it correctly.
A reactivity-prediction or drug-design ML or cheminformatics role (though you will work closely with colleagues in these areas).
A role with a playbook. Little of this has been done at our throughput, and you will be the first person at Satomic doing it full time.

Expected Outcomes

By day 30, you will have:

Working command of our analytical stack: the instruments, the acquisition methods, the existing processing pipeline, and where it currently fails.
Quantified the false-negative and false-positive rate of automated outcome calls on a real campaign, and named the dominant error modes with evidence.
Taken ownership of the retention time prediction prototype and said clearly what it needs to become a production dependency.

By day 90, you will have:

Delivered an automated interpretation pipeline that runs unattended across full campaigns, producing labeled outcomes with calibrated confidence and routing only genuinely ambiguous wells to human review.
Defined the label contract with the reactivity ML team: what is emitted, what uncertainty accompanies it, and which wells are excluded and why.
Put retention time prediction into production against two uses: flagging co-elution risk during building block selection, and qualifying which samples can share an injection.
Made internal standards work at scale, including the case where the standard changes from plate to plate.

By year 1, you will have:

Cut manual analytical review to a small fraction of wells, holding or improving label quality while throughput grows.
Built the first model that predicts CAD response from chemical descriptors, trained on ground truth we generate in house.
Delivered high-throughput feature annotation that putatively identifies side products and impurities across a plate, and flags recurring unexpected products as leads for new reactions and condition sets.
Established how we measure label quality itself, including the ground-truth experiments that keep the models honest as chemistry expands.

By year 5, you will have:

Built the analytical data science function at Satomic and the team that runs it.
Made standard-free quantitation reliable enough that analytical throughput stops constraining how fast the platform generates data.
Used raw instrument data to explain reaction mechanisms as well as outcomes.
Published work that moves analytical chemistry data processing forward, using data at a scale few groups have.

Qualifications

PhD +4 years, MS+7 years, BS +10 years in chemistry, analytical chemistry, data science, or related field.
Real fluency in analytical chemistry, particularly chromatography and mass spectrometry. Untargeted or semi-targeted experience, including building or running the data pipeline for a modern metabolomics or exposomics lab is a strong signal.
Experience processing raw chromatographic and mass spectrometry data: you have written or tuned code for peak picking, baseline correction, deconvolution, or feature detection, and you have found and fixed errors in an analysis pipeline.
Experience working directly with raw vendor files, and an understanding of what is lost when data is exported as a peak table. This includes analog detector channels (CAD, UV) that software typically handles poorly, and moving that data into tools that can run in production.
Working knowledge of charged aerosol detection (CAD), how response factors behave, which analytes it measures poorly.
Automated analysis that runs unattended at scale. You have built pipelines that process thousands of samples without manual triggering, and you have scaled one through a large increase in volume.
Interest in predicting analyte behavior from structure. Detector response, ionization efficiency, and retention time can be learned from structure, but no one has done it reliably for diverse compounds. This is the most open space in the job.
Cheminformatics and data management exposure. RDKit or equivalent for structures, descriptors, and property prediction. Experience designing and populating databases of analytical results that link raw data, processing parameters, and outputs to the sample and batch they came from. Experience writing queries that others use to pull from those databases. SQL fluency expected.
Rigor about calibration and uncertainty. A label we train needs an error estimate.
Ability to specify software precisely without building it. Python fluency is assumed and you will prototype, but the most of your impact will come from deciding what our analytical tools should do and getting engineers and coding agents to build them.
Nice to have: molecular networking or MS/MS annotation of unknowns, NMR and GC data analysis, recent bench experience, synthetic organic chemistry familiarity, familiarity with Agilent and Waters LC-MS and their data formats (ie .d files)

Overview of Satomic’s Interview Process

Hiring Manager Call (15 minutes): The hiring manager will contact you to schedule a 15 minute call to give an overview of Satomic and the role and better understand your career goals, experience, and motivations.
Technical Screen (30 minutes): Selected candidates will have a 30 minute technical discussion with an advisor about their experience with analytical chemistry, measurement science, and data pipelines.
Cores Interviews (4x45 min): The hiring manager will coordinate a block for you to have four 45 minute back-to-back interviews with different members of the Satomic team, who will ask questions about your past experience and fit for the role.
Final round (90 minutes): Candidates that are selected for the final round will be given a scoped take-home assignment representative of the work at Satomic. The candidate will be asked to present on their past work most relevant to the open position (45 minutes), along with a technical review of their take-home submission (45 minutes).

Compensation

This role offers a base salary of $165,000-$180,000, along with meaningful equity ownership and competitive benefits.

We are an early-stage company and design offers to balance cash compensation and equity ownership. The salary range above reflects base pay only, and we work with candidates to build a package that aligns with their preferences. Final offers are based on a candidate's experience and expected impact. For candidates exceptionally aligned with the role we are comfortable positioning offers toward the top of this range and beyond.

Diversity & Inclusion

We know there’s a serious lack of diversity in our industry, and that needs to change. At Satomic, we’re committed to helping drive that change. Our culture is built on inclusion, humility, humanity, and ingenuity: values that guide how we work with one another and how we approach the challenges of scaling science. We believe that bringing together people with diverse perspectives, experiences, and ways of thinking is not just the right thing to do, but makes us stronger: we want every member of our team to feel they belong and can do the best work of their career here.

Our mission is to close the gap from idea to molecule. That mission carries responsibility: the chemistry we enable can accelerate access to life-changing medicines and technologies, but it can also pose risks if used carelessly. We’re not agnostic to how our platform is applied. We care deeply about ensuring it is used for good. Building a team of diverse and thoughtful voices gives us the best chance of delivering a platform that not only advances the future of chemistry and drug discovery, but does so responsibly and with lasting positive impact on our world.