Careers at MolTrace
If you've worked in regulated science and been frustrated by the toolchain, you'll feel at home. The bar is high, the compensation matches, and the work outlasts you in regulatory ledgers.
Why join MolTrace
FDA Jan 2025 AI framework, EMA reflection paper, ICH Q2(R2), ALCOA+ — these aren't boxes we tick. They shape the design. You'll learn pharma data integrity at depth most engineers never touch.
We don't ship AI defaults until they clear strict statistical validation. New detection backends launch as opt-in, get measured against an expert-curated corpus, and only become default when they cross a predefined validation threshold. If you've watched AI startups ship and patch, this is designed to feel like sanity.
Bruker FIDs, Agilent-Varian acquisitions, real LC-MS feature tables, HRMS exact masses, MS/MS fragmentation. No synthetic-only demos. The platform fuses these as one evidence stack with cross-modal contradictions surfaced as first-class warnings.
Every existing endpoint and regression test is designed to stay green as new evidence layers land. We document our ships and our validation methodology. You'll see your work referenced in changelogs and white papers within weeks.
What you'll actually work on
These are illustrative of the kind of work we ship — drawn from capabilities built across recent release cycles, not hypothetical roadmap items.
ML systems
Expert NMR references count chemical environments; peak-pickers find multiplet lines. Build the algorithm that bridges them, validate it against a fixed reference corpus, and clear a strict median-error promotion gate.
Backend · Python
Profile a dense ¹³C-spectrum hot path. Find the bottleneck without changing the public response envelope or any audit ledger entry. Re-bench against the regression corpus and ship the gain by fixture_id.
Frontend · TypeScript
Two detection backends return different per-peak shapes (GSD's open `metadata` dict vs legacy's typed top-level fields). Design the adapter + unified envelope so a single React component renders both, columns adapting to whichever fields the backend populated.
Scientific computing
Expose reduced χ², RMSE, FWHM, S/N, and baseline σ as a green / yellow / red traffic light per peak. Catch the case where one detector reports raw signal-domain values while another reports normalised — and surface the units mismatch instead of papering over it.
Validation infrastructure
Generate per-fixture A/B JSON between two detection backends on a 20-fixture corpus. Wire it into CI so any > 50% drift on any single fixture surfaces with the failing fixture_id called out by name.
Product · Regulatory
Close the loop between a detected impurity, a residual-solvent class, an ICH Q3D limit, and a dossier-linked action item — with the audit ledger entry preserved every step of the way.
How we work
Our four design commitments don't stop at the codebase. They shape how we run meetings, how we review PRs, and how we decide what to ship.
Every endpoint, every regression test, every white-paper claim must stay green across releases. New evidence layers ship alongside old ones; we don't break the contract.
The technical white paper is part of the product, not a separate deliverable. Every important change updates the relevant section in the same task that ships the change.
ML engineer + analytical chemist + regulatory affairs reviewer on the same feature. The combinations are the moat.
Have a view. Then publish the corpus, the gate, and the threshold that would change your mind. We pre-register validation methodology before shipping.
Our hiring philosophy
We test for
We won't waste your time on
Open roles
We don't post placeholder roles to look like we're hiring. When a category is empty, register interest — we route shortlisted intros directly to the hiring manager when the role does open.
Backend (Python · FastAPI), Frontend (TypeScript · Next.js · React), ML systems, infrastructure, data integrity.
Analytical chemistry, NMR / LC-MS / HRMS / MS/MS, peak categorisation curation, validation-corpus design.
FDA AI framework + EMA reflection paper alignment, ICH Q2(R2) audit-ledger expertise, dossier-template authoring.
Sales (pharma R&D + CRO), customer success, partnerships, scientific marketing.
Compensation & benefits
The benefits below are the ones we picked because they make the work better. The standard health / parental / retirement floor is covered too — it's the floor, not the differentiator.
Conference + publication budget
ACS, Pittcon, RSC-ICMS, ENC, ASMS, plus your travel + paper-processing fees. We want you publishing what you build.
Scientific compute, not just dev laptops
Real budget for retraining shift-prediction models, running corpus expansions, and benchmarking on instrument data — not just running unit tests.
On-instrument time
Site visits to labs we work with as we onboard them. The platform feels different once you've watched an analyst use it on real instrument data.
Remote-friendly within regional timezones
Most roles are designed to support hybrid or remote within Americas / EMEA / APAC. We anchor synchronous work to your timezone, not a single HQ.
Transparent bands, geo-parity within seniority
Salary bands published at offer time, indexed annually. Equity meaningful at every level. Region parity within seniority — we don't discount EMEA / APAC offers below market.
Health, parental, retirement — covered as the floor
Comprehensive medical / dental / vision by region, 16 weeks paid parental leave, retirement match. Standard expectations; not a perk.
Interview process
30 min
Hiring manager. We share the role spec, the team you'd work with, the actual problems open. You share what you'd want to build. No screen-share, no surprise quizzes.
60 min
A specialist on the team. We walk through a real shipped artifact (a Phase release, a validation gate, a regulatory dossier template) and you tell us what you'd change, what you'd measure, what you'd test.
4-6 hours
A real PR-style task scoped to a few hours. We pay for your time. You get the same spec a teammate would get and you can ask clarifying questions throughout. No artificial time limit.
Half day
Three short conversations with future peers — one technical, one cross-functional, one with someone whose work yours would block on. We ping a reference of your choice. Decision within five business days.
Hubs & remote
We anchor synchronous work to your timezone, not a single HQ. We're building presence in regions chosen for proximity to pharma R&D and regulatory ecosystems.
Americas
Where we're building presence
Remote within ET / CT timezones · Hybrid options as we open space
EMEA
Where we're building presence · Regulatory liaison
Remote within GMT / CET timezones · Hybrid options as we open space
APAC
Where we're building presence
Remote within SGT / JST timezones · Hybrid options as we open space
Tell us what you'd build. We'd rather hear what you'd build than match you to a posting.