About Tonic
Tonic builds the data infrastructure behind modern AI. We generate the synthetic environments that agents are trained and tested in, and we de-identify real enterprise data so it can be used safely in training and evaluation. Eight years in, we work with frontier AI labs pushing the edge of what models can do, and with hundreds of enterprises including Fidelity, Comcast, eBay, and Vanguard, on the data problems that sit at the center of where AI is going next.
About The Role
The models you build here are load-bearing. The environments you generate decide whether an agent is ready to ship or only looked good in a demo. The synthesis and de-identification models you train decide whether a bank can safely put its data near a model at all. And the work spans real range: in one week you might build evaluation that separates the best models from the rest on real tasks, train a synthesis model where both fidelity and downstream utility have to hold, and improve entity detection on messy production data. Real enterprise data, real stakes, and problems that don’t have textbook answers yet.
What You'll Do
Design and build the systems that generate longitudinally coherent synthetic environments for agent training and evaluation, including persona modeling, task generators, and verifiable ground truth.
Build and maintain synthesis models that generate realistic replacement values at very large scale, preserving format, statistical distribution, and semantic consistency so de-identified data stays useful downstream.
Train and improve the NER models behind our entity detection, driving accuracy and recall across free text, structured fields, and mixed enterprise data at scale.
Build evaluation infrastructure that grades agent outcomes, not just traces, and produces real discrimination between frontier models on real tasks.
Fine-tune and evaluate open-weight models on Tonic-generated data, and turn benchmark results into product and research direction.
Expand coverage into new domains, languages, and entity types, and handle the long tail of formats and edge cases that real customer data throws off.
Own model evaluation across the board: precision and recall on detection, utility preservation on synthesis, and outcome-level grading for agents.
Optimize inference so models run efficiently on large volumes of sensitive data inside customer environments.
Partner directly with frontier labs and enterprise ML team to turn hard data problems into shipped model improvements.
What You’ll Bring
5+ years (or PhD with 1+ years) building production ML systems, with real depth in some combination of LLMs, agents, RL, NER, or information extraction.
Hands-on experience training and shipping models to production, and a pragmatic bar for quality: you know how to measure it, where it breaks, and when it's good enough to ship.
Experience with generative or synthesis models where output fidelity and downstream utility both matter, not just plausibility.
Strong software engineering fundamentals. You write code others build on.
Fluency with modern training and eval stacks (PyTorch, distributed training, standard agent and benchmark frameworks).
Comfort working with messy, sensitive, real-world data and the privacy constraints that come with it.
A track record of framing ambiguous problems and driving them to measurable, shipped results.
Bonus: synthetic data generation, data privacy or de-identification, or benchmark construction.
Benefits We Offer
Competitive salary and equity
Unlimited paid time off
401k plan with employer contribution
Medical, dental, and vision insurance
Generous parental leave policy
Remote-friendly work environment
Tonic.ai has been de-identifying sensitive data since 2018, and hundreds of enterprises use our products, among them eBay, UnitedHealth Group, Qualcomm, Commonwealth Bank, and Mercor. Our newest line of business takes that work somewhere it hasn't gone before: company operating data. Email, chat, tickets, documents, code, CRM and HR records are some of the most valuable training data left for AI labs, and almost none of it can change hands until the people in it are protected.
We sit between the company that owns the data and the lab that wants it, as the independent third party. We take in the corpus, find the people and organizations referenced across all of those systems, give each real identity one consistent synthetic replacement, rebuild every file in its original format, and deliver the result with a written report on how it was done. The data still reads like a company doing its work, which is what makes it worth training on.
Demand is running ahead of our capacity to deliver. Labs are buying, data owners are selling, and the buyers want datasets in days to weeks, not quarters. The technology can meet that timeline. What turns it into a delivered dataset is an engineer who can take a messy, half-documented corpus from a new client, get it through the pipeline, catch what breaks, and ship something both sides will sign off on.
That's this job. You’ll be one of the first engineers dedicated to this business, working directly with our technical founders, data science team and platform engineers, and getting close to the frontier AI labs that are the end buyers of this data.
Own the delivery of client de-identification projects, from the first look at the raw data to a dataset the client accepts.
Run each corpus through our de-identification pipeline and check the output for accuracy: missed entities, wrong replacements, identities merged or split incorrectly, files that won’t rebuild.
Diagnose problems as they appear. Where you can, fix them for the whole solution or for the one case. Where you can’t, understand the issue well enough to explain it clearly to the engineer who can.
Rerun the pipeline, confirm the fix worked, then move on to the next step, which often includes working with our annotation team to validate quality.
Build and adjust the ingestion and transformation steps that turn raw exports (mailboxes, chat archives, document dumps, ticketing systems, databases) into processed, ready-to-use datasets.
Write code that gets merged, within the guardrails of our existing platform, including custom handling for client-specific one-offs.
Run reconciliation checks so file counts and content tie from ingest to delivery.
Contribute to the technical sections of the client report: what was treated as sensitive, what was kept and why, how each file type was handled, and known limitations.
Walk clients through what was done and the resulting data, alongside Tonic’s founders at first and on your own as you grow.
Turn every one-off fix into something reusable: documentation, runbooks, and improvements to the product.
A bachelor’s degree in a technical field (computer science, engineering, math, statistics, physics, bioinformatics, or similar).
Several years of hands-on work in data engineering, backend engineering, or a solutions/forward-deployed role where you shipped production systems for customers.
Strong experience in Python and other backend programming languages, plus SQL. Comfortable reading and debugging other people's code.
Experience building repeatable pipelines that process large volumes of structured, semi-structured, and unstructured data. You understand how data flows through a system and can tell when output is wrong.
AI-native workflow. AI tools and agents are central to how you work, not an add-on. You know where they’re reliable, where they need your oversight, and you don’t trust their output without checking it.
Self-direction. You'll often be the only engineer on an engagement, and you'll need to make sound calls without waiting for someone to tell you what to do.
Clear communication with very different audiences. You can walk a client's engineer through a pipeline failure in the morning and explain a tradeoff to a senior executive in the afternoon.
A record of delivering complex projects on short timelines, measured in days and weeks.
Experience with model inference, ML pipelines, or PyTorch.
Experience building or maintaining data masking, de-identification, ETL or synthetic data tooling.
Interest in working close to frontier AI labs and the training-data ecosystem.
Experience at an early-stage company or on a new product line.
In your first month, you'll learn the platform and the delivery process and run part of a live engagement alongside the team. By the end of your first quarter, you'll be running engagements on your own, from scoping to delivery, and the runbooks and tooling you've written will have made the second and third ones faster than the first. After that, you'll help decide how this function grows, including who we hire next.
You’ll be early in a fast-growing business solving a problem the AI industry doesn’t yet have a good answer for.
You’ll work alongside technical founders, with a direct line from what you see in the field to what gets built.
You’ll get close to the frontier AI labs and build experience at the center of the training-data market.
The work matters beyond any one deal: getting company data to labs without exposing the people in it is a problem the whole industry has to solve.
Competitive compensation and equity.