MISSION

We are on a mission to democratize access to financial intelligence. Every company should have financial insights at a fingertip – not just big enterprises with dedicated finance teams. For the first time, AI makes this possible. In the intelligence age, business owners with no financial expertise will get access to all the latent knowledge hidden in their books. Businesses will operate with a level of clarity, confidence, and speed that is almost unimaginable today.

Tabula is the accounting software that does the work for you. From paying, to getting pay; organizing expenses, generating insights and filing taxes. By automating the manual work, we give SMBs easy access to financial insights and let them focus on what they do best: running their business.

TRACTION

Backed by Y Combinator and raised a $4.8m Seed round from LocalGlobe. Since openly launching this summer we have seen strong traction with > 500 companies onboarded, $280m in expense volume processed and 10% weekly growth.

ML PROBLEMS WE WORK ON

1) Vast Problem Space – We are automating highly complex workflows end-to-end, consisting of many individual steps and decisions currently performed by trained experts. Some examples include:

• Financial Insights: Using AI code generation to model financial data, according to the custom needs of each company. Instead of static dashboards or convoluted excel sheets, companies get tailored insights.
• Document Extraction: Most data inputs in finance are still unstructured, such as invoice PDFs. The challenge: extract dozens of data points with near-zero error, at a scale of several tens of thousand pages per day 👉 Achieving >99% Extraction Accuracy
• Expense Classification: Every item a company purchases must be mapped to the correct expense category. Getting this right requires a deep understanding of the nature of the purchase and the business context.
• Bank Reconciliation: Transactions must be matched against invoices and receipts with high reliability, even in cases of incomplete, ambiguous, or conflicting data.

To build the best in class AI, we have accumulated large, expert-labeled datasets consisting of several hundred thousand invoices and transactions.

2) Confidence / Interpretability – Our customers are criminally liable if they make major mistakes in their finances. If we only reach 95% accuracy on a given automation, customers will still review 100% of outputs to catch the 5% of errors. This review process erodes much of the value of automation. We are therefore heavily investing in confidence calibration, LLM interpretability, and hallucination safeguards to ensure that automation can be trusted at scale.

3) From agentic LLMs to traditional ML – for accountants Tabula feels like Cursor does to devs. Achieving this requires a broad technological stack, from closed-source LLMs to custom-trained open-source models tailored for financial workflows.

4) Research Partnerships – We actively collaborate with leading AI research groups. This includes an alpha partnership with OpenAI on reinforcement learning for frontier reasoning models, and a collaboration with researchers from the Cambridge Van der Schaar Lab.

WHO THRIVES HERE

• You live and breathe code — it’s the first thing on your mind in the morning and the last at night. You get energy from diving into new technologies.
• You love challenges – we seek out hard problems. Every painful technical barrier we cross, puts another 100 miles between us and the legacy incumbent.
• You care about users – code only matters if it changes their world. Over-engineering is fatal in early-stage. But the best teams do both: move fast and ship great code.