Research Scientist, Analytical Chemistry

About the company

Onepot AI leverages machine learning and an automated laboratory to rapidly synthesize complex chemical molecules for drug discovery partners.

About this role

onepot is automating chemistry. Our goal is to enable a self-improvement loop for chemistry by combining AI and advanced robotics. In this loop, AI systems design experiments, robotic systems execute them, and the resulting data improves the next generation of models. Analytical chemistry is the feedback signal of that loop. Every reaction the platform runs ends as a measurement, and those measurements are what our models learn from — which makes this the role where the platform's notion of what worked gets defined. We are building a small, unusually ambitious team and are looking for an analytical chemist who wants to own that definition. WHY THIS IS DIFFERENT HERE Most analytical work optimizes one method for one molecule. Ours has to hold up across a platform that runs broad, structurally diverse chemistry in parallel — compounds that ionize strangely, products that hide behind adducts and in-source fragmentation, regioisomers that share an exact mass, detectors that saturate, peaks that are not what they claim to be. Interpretation has to work at scale, and it has to be right. And the data has two customers: people and models. A number that leaves the building goes to a customer with our name on it, and the same number becomes training data for the next generation of our models. No one has built an interpretation layer that holds up across chemistry this diverse — the tools do not exist off the shelf, and building them is most of why this job is interesting. THE ROLE You will own analytical data quality end to end — the methods, the interpretation standards, and the pipelines that apply them to everything the platform runs. Day to day, you will take on the interpretation problems that automation cannot yet handle: assigning fragmentation that no library contains, resolving ambiguous identity, deciding whether a marginal result is chemistry or artifact. You will decide when a question needs orthogonal evidence — NMR, accurate mass, a separation we do not run yet — and go get it. If the platform is missing a technique or a method, whether that is SFC, better use of NMR, or something we have not thought of, we expect you to evaluate it, bring it up, and wire its data into our systems. The ceiling of this role is not how many samples you can look at yourself. We expect you to turn your judgment into things that run without you — decision rules, scripts, models — so that your standards apply to every well the platform produces, around the clock. We build and use AI agents heavily, including for first-pass analytical interpretation, and you will work with them and with our ML team to push more of the interpretation into software. Work here is finished when anyone can use it: a tested module, an endpoint, a button in the app — not a notebook, and not a ritual only you can perform. The routine parts are automated, or will be — you are hired for judgment, and for building the things that scale it. We are hiring at the top of this craft, and the technical bar is the point: the platform multiplies whatever judgment we build into it, so it has to be the best available. EDUCATION - Ph.D.-level depth in analytical chemistry or a related field, or the demonstrated equivalent. We care more about what you have solved and built than about the credential itself. EXPERIENCE - A track record of solving structures others could not: unknowns without reference standards, unexpected products, trace impurities — with the orthogonal evidence to prove the assignment. - Experience taking an analytical capability from an initial need to something that runs routinely — method, instrument, and data flow. - Experience interpreting analytical data in volume, where the throughput forced you to systematize your own judgment rather than eyeball everything. - Evidence that you automate yourself out of repetitive work — scripts, decision rules, or models that replaced something you used to do by hand. - Experience in a startup, research group, or other environment where you had significant ownership and limited resources is particularly relevant. SKILLS - Expert small-molecule mass spectrometry. You rationalize fragment series mechanistically — charge-driven versus charge-remote pathways, characteristic neutral losses — and defend an elemental composition from mass defect, isotope fine structure, and ring-and-double-bond logic rather than a library hit. You know what electrospray does to different chemotypes, and how to pull quantitative signal out of a detector past its linear range. - Deep NMR: routine command of the 2D suite (COSY, HSQC, HMBC, NOESY), de novo assignment of unfamiliar scaffolds and mixtures, relative stereochemistry from NOE and coupling analysis, and quantitative NMR against an internal standard. Heteronuclear work (¹⁹F, ³¹P, ¹³C), variable-temperature experiments, in situ reaction monitoring, and DOSY on mixtures should be tools you have actually reached for, not terms you recognize. - Separation science at the method-development level: stationary-phase and mobile-phase selection from first principles, gradient design, and chiral or SFC methods when a problem calls for them. - Quantitation and statistics you can defend: calibration design, internal standards, ion suppression and matrix effects, limits of detection — and replicates, noise floors, and distributions rather than single traces. - Fluency with AI agents and modern AI tooling, and a habit of prototyping quickly. We expect successful candidates to already work this way. - Obsessive attention to detail. A wrong number here does double damage: it misleads a customer and it trains a model. PARTICULARLY RELEVANT EXPERIENCE Experience in any of the following would be useful, but we do not expect one person to have all of it: - Impurity and degradant identification, or structure elucidation of unknowns at trace level - Ion mobility, multi-stage MS, or other advanced acquisition strategies - Two-dim