# Pulse

Aiolum story: https://aiolum.com/story/pulse/ · added 2026-11-12 · updated 2026-11-12

**In short.** Pulse (Document extraction API for LLM pipelines, USA): Self-serve plan with usage pricing, November 2026. Evidence grade E: no revenue figure was checked. Team: 2 founders in 2024; 22 people on the Y Combinator page, read in November 2026; US$3.9M seed round, per the company. Launched July 2024; still running.

Document extraction API for LLM pipelines · USA · AI-native · started small, grew after funding · still running · group: Documents, data and slides

- Outcome: Self-serve plan with usage pricing, November 2026 (evidence grade of the headline number: E; no revenue figure)
- Launch: July 2024; months from launch to the outcome: 28
- Team: 2 founders in 2024; 22 people on the Y Combinator page, read in November 2026; US$3.9M seed round, per the company
- Outside funding: Seed round of US$3.9M led by Nat Friedman and Daniel Gross, with Y Combinator and others (company blog); a Business Insider headline says US$3 million
- Main factor: Own problem
- Meters, 1 to 4: evidence strength 1, copyable pattern 1, capital needed 3
- Tags: AI-native, Small team, Funded, Running, B2B, Documents

Links:

- Visit runpulse.com: https://www.runpulse.com/

**The product.** A service that turns PDFs, scans, images and spreadsheets into structured data that language models can read. A program sends a file to Pulse through an API and gets back text, tables and the position of each word on the page.

**Why it worked.** The founders built the tool first for a supply chain use case and found no other product that worked. They planned to keep it in-house. Requests from dozens of companies changed that plan, according to the company's Y Combinator launch post.

Evidence grades: A = disclosed by a public company, a filing or a payment processor, or verified by one; B = two or more independent reports agree, or a journalist confirmed it; C = a founder claim that no one else verified; D = reported, not verified: one founder claim or a third-party estimate with no second source; E = no revenue figure found, so the headline is a non-money fact and the other facts carry their own grades; calc = computed on this page from graded facts.

## Summary

Pulse sells document extraction: software that reads a document and returns its content as data. It uses OCR (optical character recognition, which reads printed or handwritten text from an image) and models that find tables and charts. A program calls it through an API, an interface that one program uses to send work to another. The price unit is a credit, which the pricing page ties to the pages processed.

Two founders started the company in 2024, Sid Manchkanti as chief executive and Ritvik Pandey as chief technology officer. The Y Combinator page lists it in the Summer 2024 batch with 22 employees. The company's blog says a seed round of US$3.9M was led by Nat Friedman and Daniel Gross. The Y Combinator page also lists a Business Insider headline that gives US$3 million.

No revenue figure was found, so this story is grade E. The evidence for sales is the overage price in the company's pricing page data and documentation, and its billing terms. Two customer posts that the company wrote itself add to it. No outside source confirms that any customer pays.

## Why did Pulse work?

### It began as an in-house tool for one supply chain use case

The Y Combinator launch post says the founders built the API for supply chain documents. It says they could not find another product that worked. They planned to keep it in-house. Requests from dozens of companies led them to offer it. This is the company's own account.

### A free start of 20K pages replaced a demo-only offer

The July 2024 copy of the site offers only "Book a demo". The posts of January and February 2026 open the product to anyone, with up to 20K pages free. Beyond that, the pricing page data lists US$0.015 a credit. Nothing shows how many users moved to paid use.

### The company shows its case through customer posts and benchmarks

The blog carries two customer posts, on Fountain and Syntra, and the home page banner points to a table-extraction benchmark. In the Fountain post, the company says processing time fell by 90%. These are the company's statements. Fountain's quoted team lead calls the product "more cost-effective" than other tools.

## How it was built

These details come from the company's own pages and posts. They are grade C.

- The launch post says the founders trained their own vision language models and OCR methods. It says most competing tools wrap Amazon Textract or Gemini.
- The home page lists five steps: layout detection, OCR, reading order, table structure recognition and fine-tuned models for charts and figures.
- The seed post says the product combines schema mapping with fine-tuned extraction models. The company's pages offer cloud use, and private cloud or on-premises use for the Enterprise plan.
- The company's site is built with Webflow, and its careers page embeds the Ashby job board.
- The home page says its security is verified by Oneleet and independent audits. No audit report was read.

Sources: [3, 1, 4, 2, 5]

## Who tried this before Pulse, and why did they fail?

Three companies in our database of failed startups sold document, lease or spreadsheet software to business teams. The values below are from that database and are not independently verified.

The failed startups come from our database of dead startups, retold here in our own words; Its entries are AI-assisted summaries of public sources, so they carry no evidence grade.

### Docucharm (2019 to 2022; funding US$3.0M; cause: Competition)

- Promise: A cloud service to digitize, organize and find business documents for operations teams.
- Why it died: The record names Google and Microsoft, which offered document tools built into their larger suites. It also names high costs of digitization and storage.
- Record: 

### Acre Platforms (2018 to 2025; funding US$12.0M; cause: No market need)

- Promise: One platform for commercial lease management, document automation and deal coordination between landlords, tenants and brokers.
- Why it died: The record says the pain was not acute enough for users to change habits. It says each business client needed heavy customization and a 3 to 6 month setup.
- Record: 

### DataNitro (2012 to 2015; funding US$120K; cause: Technology shift)

- Promise: Run Python scripts directly inside Excel for data analysis and automation.
- Why it died: The record says the product was limited by Excel's design and needed users who knew Python. It also says later Excel versions added Python.
- Record: 

## What changed

The three predecessors sold document or spreadsheet work to business teams. Their records name large platforms that added the same features, and customers that needed custom setup. Pulse works in the same area and faces the same two risks.

Two things differ. Pulse sells an API that other software calls, and the price is charged per credit, so setup for each customer is smaller. The company also says it trained its own models. No source shows how many customers pay or whether large platforms will match the quality.

## Can one person copy Pulse now?

**Verdict: The idea is copyable. The model work is not, for a solo founder.**

A small team can sell a narrow document service when it first built the tool for its own work. The company's own account shows the order: an in-house tool, then demand from outside, then a public API. The sources show no revenue, so a copy should test payment before building models.

1. Pick one document type that you already handle at work, such as supply chain paperwork.
2. Offer the tool as an API with a free allowance, and charge per page beyond it.
3. Publish customer posts and test results, since buyers compare tools on accuracy.
4. Add private deployment only for customers who ask, as the Enterprise plan does.

**What is still open.** Other document types that existing tools handle badly, such as handwritten forms, engineering drawings and scanned maps.

## How much revenue does Pulse make? The numbers, graded

A two-founder company from 2024 that grew to 22 people and sells usage-priced document extraction. No revenue figure was found, and the company states every row except the Y Combinator ones

| Fact | Value | Grade | Note | Sources |
|---|---|---|---|---|
| Legal entity | Pulse Software, Corp. | C | The terms of service (last revised 1 December 2025) and the privacy policy (last updated 1 December 2025) name this company, in San Francisco. | [9, 10] |
| Company start | 2024 | C | The Y Combinator page says "Founded: 2024" and "Batch: Summer 2024". A copy of the company's older site, trypulse.ai, from 17 July 2024 carries "© 2024 Pulse Software, Corp." and the name "Pulse AI". | [3, 11] |
| Earlier name or product | none found; same company as "Pulse AI" | C | The July 2024 copy offers "Book a demo" only, and one review on it calls the product "PO automation software". The launch post on the Y Combinator page says the API was first built "for a very specific use case in supply chain". | [11, 3] |
| Founding team | 2: Sid Manchkanti (CEO) and Ritvik Pandey (CTO) | C | The Y Combinator page says "Founded in 2024 by Sid Manchkanti and Ritvik Pandey". The seed post says "Pulse was founded by Sid Manchkanti (CEO) and Ritvik Pandey (CTO)". | [3, 4] |
| Team size today | 22 | C | The Y Combinator page says "Team Size: 22" and "Pulse has 22 employees", with no date. The page also lists 9 open jobs. | [3] |
| Funding | seed round of US$3.9M | C | The company's blog says it was led by Nat Friedman and Daniel Gross. The Y Combinator news list shows this post dated 19 February 2025, but the post page itself shows 17 December 2025. | [4, 3] |
| Press headline on the round | US$3 million | C | The Y Combinator page lists a Business Insider headline, "AI Startup Pulse Raises $3 Million From Nat Friedman and Daniel Gross", dated 19 February 2025. Business Insider blocks automated reading, so the article was not read. | [3] |
| Overage price | US$0.015 a credit beyond 20K free credits | C | The pricing page's structured data says "Overage billed at $0.015 per credit". The visible table shows "Custom" for the overage per credit. The company's documentation says pay-as-you-go use costs "$0.015 per credit", and tells an account with no payment method to add one in the dashboard. | [2, 13, 14] |
| Billing terms | recurring subscriptions, charged to a stored payment method, no refunds | C | The terms say customers who buy paid services "agree to pay us the applicable fees", renewals are automatic, and payments are "nonrefundable". | [9] |
| Named customers | Fountain and Syntra | C | The company's own posts of 4 March 2026 (Fountain) and 26 January 2026 (Syntra) describe them as customers. Neither post states a price paid. | [5, 6] |
| Pages processed | 600 million (9 January 2026) and over 1 billion (2 February 2026) | C | Company statements in two blog posts. The home page repeats "over 1 billion pages". | [8, 7, 1] |
| Time from the first dated public page to the headline | 28 months | calc | Calculated: July 2024 to November 2026. | [11, 2] |

How to read this table. Grade C is a statement from the company's own pages or from its Y Combinator profile. A "calc" value was computed on this page. No row is grade A or B.

## Pricing

Pulse pricing read from the company's pricing page on 12 November 2026 (grade C)

| Plan | Per month, billed yearly (US$) | Per month, billed monthly (US$) | Limit |
|---|---|---|---|
| Self Serve | 0 | 0 | 20K credits free; US$0.015 a credit beyond (page data and documentation) |
| Enterprise |  |  | custom contract |

The page shows a Self Serve plan with 20K free credits. Its visible table gives the overage per credit as "Custom". The page's structured data gives US$0.015 a credit, and the company's documentation gives the same price for pay-as-you-go use. The page source holds a switch for monthly or annual billing with 15% off, in a block marked hidden, with no amount. Enterprise has custom pricing.

## When did Pulse launch, and how did it grow? Timeline

From a demo-only site in July 2024 to a free sign-up and usage price in 2026. Each entry carries its grade.

- **2024** (grade C [3]): Y Combinator lists the company as founded in 2024 and in its Summer 2024 batch, with two founders.
- **17 July 2024** (grade C [11]): The first dated copy of trypulse.ai shows "Pulse AI" with a "Book a demo" button and the footer "© 2024 Pulse Software, Corp."
- **19 February 2025** (grade C [3, 4]): The Y Combinator news list dates the announcement of the US$3.9M seed round to this day. It dates a Business Insider article, with a figure of US$3 million, to the same day.
- **9 January 2026** (grade C [8]): The company's post says the platform is open to sign-up and that it has processed more than 600 million pages.
- **2 February 2026** (grade C [7]): A post says one billion pages were processed. It says anyone can use Pulse and process up to 20K pages for free.
- **12 November 2026** (grade C [2, 3]): The pricing page shows the Self Serve plan with 20K free credits. The Y Combinator page shows a team of 22.

## What could go wrong for Pulse?

- **What is not verified.** No revenue figure was found, so this story is grade E. The evidence for sales is the company's published price of US$0.015 a credit, its billing terms, and two customer posts that the company wrote. The visible pricing table shows "Custom" where the page data gives that price. No source states that a customer has paid. A reader could look for a payment-processor profile, a customer interview or a funding article that states revenue.
- **Trust signals the sources do not support.** The home page shows five testimonials from unnamed roles, such as "Top 10 PE Firm", and one from a named Syntra executive. It shows a row of logos, among them Samsung, Cloudera and UC Berkeley, and no source confirms them as customers. The page speaks of "Fortune 10 enterprises" and names none. It shows SOC 2 Type II, ISO 27001, GDPR and HIPAA badges, verified by Oneleet, and no report was read. The count of 1 billion pages is the company's own. The seed post is dated 19 February 2025 by Y Combinator and 17 December 2025 by the post page. The round is US$3.9M in the company's post and US$3 million in a headline.
- **The team size has no date.** The Y Combinator page says 22 employees and gives no date for the count. The company's seed post calls itself "a small but fast-growing team of engineers". The 2 founders started it in 2024, and this story uses 22 as the size today.
- **Agents can pay in a stablecoin.** The company's documentation says software agents can pay per request in USDC, a stablecoin, on the Tempo blockchain, with no account. The company's post of 11 May 2026 announced this. Payment by card through an account is the other path. No source shows how much use either path has.

## Sources

1. Pulse home page, read 12 November 2026 — https://www.runpulse.com/ (grade C). Read in full. Source for the pipeline steps, the testimonials, the security badges and the counter of pages processed.
2. Pulse pricing page, read 12 November 2026 — https://www.runpulse.com/pricing (grade C). Read in full, including the structured data in the page source. Source for the free credits, the overage price and the Enterprise plan.
3. Y Combinator company page: Pulse, read 12 November 2026 — https://www.ycombinator.com/companies/pulse-3 (grade C). Read in full. Source for the founders, the batch, the founding year, the team size, the news list and the launch post.
4. Pulse blog: Announcing our $3.9M Seed Round — https://www.runpulse.com/blog/pulse-seed-nfdg-venture (grade C). Read in full. Source for the seed round, its investors, the founders and the founding year. The page shows the date 17 December 2025.
5. Pulse blog: How Fountain Cut Document Processing Time by 90% with Pulse, 4 March 2026 — https://www.runpulse.com/blog/how-fountain-cut-document-processing-time-by-90-with-pulse (grade C). Read in full. A customer post written by the company. Source for the named customer and the claimed results.
6. Pulse blog: How Syntra Uses Pulse to Process $1B+ Of Medical Charges a Year, 26 January 2026 — https://www.runpulse.com/blog/syntra-pulse (grade C). Read in full. A customer post written by the company. Source for the named customer and the testimonial.
7. Pulse blog: One Billion Pages Processed, 2 February 2026 — https://www.runpulse.com/blog/one-billion-pages-processed (grade C). Read in full. Source for the one billion pages statement and the free 20K pages.
8. Pulse blog: Pulse Opens Its Platform, 9 January 2026 — https://www.runpulse.com/blog/pulse-opens-its-platform (grade C). Read in full. Source for the 600 million pages statement and the public sign-up.
9. Pulse terms of service, last revised 1 December 2025 — https://www.runpulse.com/terms-of-service (grade C). Read in full. Source for the legal entity, the subscription payment terms and the refund rule.
10. Pulse privacy policy, last updated 1 December 2025 — https://www.runpulse.com/privacy-policy (grade C). Read the first sections. Source for the legal entity and the two named representatives.
11. Internet Archive copy of trypulse.ai, 17 July 2024 — https://web.archive.org/web/20240717173504/https://www.trypulse.ai/ (grade C). Read in full. Source for the name Pulse AI, the demo-only offer and the footer dated 2024.
12. Database of failed startups, local export — /method/#dead (grade C). Record pages read on the website on the day of writing. AI-written summaries, not independently verified.
13. Pulse documentation: Pay Per Request (MPP), read 12 November 2026 — https://docs.runpulse.com/advanced/mpp-payments (grade C). Read in full. Source for the price of US$0.015 a credit for pay-as-you-go use and for payment in USDC.
14. Pulse documentation: PAYMENT_REQUIRED error, read 12 November 2026 — https://docs.runpulse.com/errors/payment-required (grade C). Read in full. Source for the instruction to add a payment method in the dashboard.

Contact: aiolum@livyden.com. Run Pulse? Send corrections or a verified figure to aiolum@livyden.com (subject: Correction: Pulse).