×
Fundamental
Enterprise Data AI
Last verified August 15, 2026 · Updated daily
What Fundamental is building
Structured data – spreadsheets, databases, transaction logs, medical records – makes up roughly 80% of all enterprise data by volume, and it's the data that actually drives consequential decisions. Which drug goes to trial. Which financial transactions get flagged. When to cut power to prevent wildfires. But LLMs, built on Transformer architectures, are terrible at this data type. They truncate rows. They lose precision over long tables. They hallucinate numbers. They can't reason over a dataset with billions of rows. For tasks where precision is the entire point, fraud detection, credit risk, clinical prediction, hallucination is not a minor limitation. It is a disqualifying failure. For the past decade, the dominant approach to extracting predictive intelligence from this data has been gradient boosting, specifically XGBoost and LightGBM, the ensemble methods that win virtually every structured data competition on Kaggle and power the majority of production ML systems in banking, insurance, and retail. These methods work well, but they require significant human labour: a data scientist must manually engineer features, handle missing values, encode categorical variables, tune hyperparameters, and rebuild the entire pipeline from scratch for every new prediction task. A fraud detection model built for one bank cannot be transferred to another without repeating this process entirely. The knowledge is in the engineer's head and the custom code they write, not in the model itself. NEXUS is Fundamental's answer to this architectural gap. It is not a Transformer. It is trained on massive quantities of tabular data across industries, it ingests raw tables with minimal preprocessing, it produces fully deterministic outputs with no probabilistic variation between runs, and it generalises across prediction tasks without task–specific fine–tuning. The claim, that a single model can be dropped onto a new tabular dataset and produce state–of–the–art predictions without the feature engineering and pipeline construction that XGBoost requires, is the central bet. Fortune 100 customers, like AWS and those in banking and healthcare already running seven–figure contracts suggest the claim is holding in production.
Why this matters
Globally, there are roughly 4 million data scientists and ML engineers, the majority of whom spend the majority of their time doing the same set of tasks: cleaning data, engineering features, building pipelines, tuning models, and maintaining the resulting systems in production. McKinsey estimated in 2023 that data preparation alone consumes 60–80% of a data scientist's working time. That is not an anecdote about inefficiency, it is a structural description of an industry where the highest–paid technical specialists spend most of their hours on work that is repetitive, manual, and in principle automatable. Jeremy Fraenkel's formulation is worth sitting with: 'If you look at what LLMs have done with unstructured data, it's been amazing. But it only covers 20% of data.' The other 80%, the tables, is where the real economic value sits. NEXUS, if the generalisation claim holds at scale, automates that majority. A data scientist working with NEXUS does not engineer features or tune hyperparameters, they define the prediction task, point the model at the data, and evaluate the output. The pipeline construction collapses into a single API call. The implication for enterprise data science teams is not subtle: the same output that currently requires a team of ten data scientists could be produced by a team of three with NEXUS handling the modelling layer. The AWS partnership is the distribution mechanism that makes this commercially significant at speed. Enterprise software sales into Fortune 500 companies typically requires an 18–month procurement cycle – vendor evaluation, security review, legal negotiation, budget approval, implementation. AWS Marketplace collapses this. Companies that already have AWS enterprise agreements can purchase NEXUS directly through their existing procurement relationship, drawing down from pre–committed cloud spend without a new vendor approval process. That is not a go–to–market convenience, it is a structural bypass of the primary barrier to enterprise AI adoption. Snowflake built a $60B+ business partly on the same insight: distribution through existing cloud relationships reduces sales cycles from years to weeks. Fundamental is starting in financial services and healthcare, the two industries with the largest structured datasets, the highest stakes for prediction accuracy, the most regulatory pressure on model explainability, and the deepest frustration with LLM hallucination. A fraud detection system cannot explain to a regulator that its model 'sometimes gets numbers wrong.' A clinical readmission risk tool cannot be deployed in a hospital that cannot audit how a prediction was made. NEXUS's determinism – the fact that the same input always produces the same output, and that output can be traced – is not just a technical property. It is a compliance feature that unlocks markets that probabilistic models cannot enter. Once embedded in a bank's fraud infrastructure or a hospital's clinical decision support system, the switching cost is effectively infinite.
Open roles at Fundamental
6 positions we're tracking. Roles are re-checked daily and removed when filled.
ML Researchers (tabular data)
First seen 4 months ago
ML Engineers
First seen 4 months ago
Enterprise Account Executives
First seen 4 months ago
Solutions Engineers
First seen 4 months ago
Data Scientists
First seen 4 months ago
Product Manager
First seen 4 months ago
Hiring outlook
7 figure Fortune 100 contracts at seed stage
Working at Fundamental
Fundamental: Enterprise Data AI. Founded , currently employees. What this means for you: opportunity to shape your role based on the company stage.
Most Fundamental jobs are based in San Francisco.
Frequently asked questions
How many jobs does Fundamental have open?
Fundamental currently has 6 open roles, last verified February 2025.
Does Fundamental hire remotely?
All current Fundamental roles are based in San Francisco.
What roles is Fundamental hiring for?
Fundamental is hiring across Data, Engineering, Sales, Product. The most recent opening is ML Researchers (tabular data).
How do I apply for a job at Fundamental?
Apply directly through the links above, or read our guide on how to actually get hired at Fundamental.
Where is Fundamental based?
Fundamental is headquartered in San Francisco.
Get Fundamental roles before they're posted
Get AI jobs like these before they hit LinkedIn. We track companies like Fundamental daily.
Get early access →