Fira Does Your Financial Research While You Go Get Coffee

The Macro: Wall Street’s Research Analysts Are Drowning in PDFs

Investment research is still, in 2025, a shockingly manual process. An analyst covering a sector might need to read through 50 quarterly earnings reports, compare revenue growth across 12 companies, cross-reference management commentary with actual financial results, and build a comp table. That’s hundreds of hours per quarter spent on what is essentially structured data extraction from semi-structured documents.

The tools analysts use are embarrassing for an industry that moves trillions of dollars. Bloomberg Terminal is powerful but costs $24,000 a year and its document search is mediocre. FactSet and Capital IQ are better for data but weak on unstructured document analysis. Most analysts end up doing what they’ve always done: opening PDFs, reading them, and manually typing numbers into Excel. In 2025.

There have been attempts to fix this. AlphaSense built a strong search product for financial documents and raised over $600 million. Sentieo (now part of AlphaSense) was doing similar work. Koyfin offers a cheaper Bloomberg alternative. Tegus built a research platform around expert interviews. But none of these tools truly automate the research process. They make it easier to find documents. The analyst still has to read them and extract the relevant data.

AI changes the calculus here because financial documents are highly structured even when they’re in PDF form. An earnings report follows predictable patterns. SEC filings have standardized sections. The data is there, it’s just locked inside formats that traditional software handles poorly. LLMs that can read, understand, and extract from these documents could compress hours of analyst work into minutes.

The Micro: A Crypto Exit and a 45-Million-User Resume Builder

Fira’s founding team has an unusual combination of credentials. Alex Tkachenko (CEO) previously founded Hashscan, a blockchain analytics company, and created Hodler, a crypto portfolio tracker he sold in 2022. Alexey Taktarov (CTO) was co-founder and CTO at resume.io, which scaled to over 45 million users before being acquired in 2021. He also authored wouter, a React router library with 800,000 monthly downloads on npm. Both have Twitter presences and are active in the developer community.

They’re a four-person team in YC’s Winter 2025 batch, working with Gustaf Alstromer. The product lets analysts delegate research tasks across multiple data sources. You can ask it to compare revenue growth across a set of companies, extract P&L data from scanned PDFs, or find every mention of a specific metric across thousands of pages of filings.

The source citation system is the technical differentiator I find most compelling. Fira cites its answers down to the specific cell, row, or paragraph. That matters enormously in finance. An analyst can’t put a number in a model and tell their portfolio manager “the AI said so.” They need to verify. Clickable citations that take you to the exact source make verification fast instead of painful.

The Excel export feature is smart product thinking. Analysts live in Excel. Any tool that forces them to work in a different environment is fighting an uphill battle. Fira reads PDFs and converts them to Excel while preserving formatting, and generates charts that look like what you’d see in an investment bank presentation. Meeting analysts where they already work is the right move.

The platform also handles internal documents alongside public filings. That’s important because investment firms have proprietary research, previous deal memos, and internal analyses that need to be searchable alongside public data. A tool that only handles public filings misses half the workflow.

The Verdict

I think Fira has one of the clearest product-market fit signals I’ve seen in a while. Investment analysts are expensive ($150K-$300K fully loaded), they spend a huge chunk of their time on repetitive document analysis, and the output of that analysis is highly structured. This is exactly the kind of work AI should be doing.

The competitive landscape is the main concern. AlphaSense is a $4 billion company with deep relationships across the buy side and sell side. They’re not going to sit still while a startup eats their lunch. Bloomberg is slow to innovate but has distribution that nobody can match. And every major LLM provider is building financial analysis capabilities into their platforms.

But Fira’s team is unusually strong for this stage. Tkachenko has built and sold fintech products. Taktarov has scaled a product to 45 million users and has serious engineering credibility in the open-source community. They understand both the financial domain and how to build products that people actually adopt.

In 30 days, I want to know how many investment firms are using Fira in their actual research workflow, not just demoing it. In 60 days, the accuracy question: when Fira extracts a number from a 10-K filing, how often is it right? In finance, 99% accuracy means you’re wrong on one out of every hundred data points, and that’s potentially a million-dollar mistake. In 90 days, the question is whether Fira can expand beyond equity research into credit, private markets, and other verticals where document-heavy analysis is the norm. The TAM is enormous if the product works. The founders have the track record to make it work.