BREAKING Ambria (YC W24) is building AI agents for the industrial world Founders Praneeth Guduguntla & Arman Rafati met as freshmen at UIUC Launched in Y Combinator's Winter 2024 batch Started as Glimmer, AI search for 10,000-page PDFs Now targeting back-office automation in manufacturing & distribution BREAKING Ambria (YC W24) is building AI agents for the industrial world Founders Praneeth Guduguntla & Arman Rafati met as freshmen at UIUC Launched in Y Combinator's Winter 2024 batch Started as Glimmer, AI search for 10,000-page PDFs Now targeting back-office automation in manufacturing & distribution
Company Founder Profile • Y Combinator W24

Ambria Wants AI Agents to Run the Industrial Back Office

The two-person YC W24 startup began life as Glimmer, a search engine for enormous PDFs. Now founders Praneeth Guduguntla and Arman Rafati are aiming their document-reasoning tech at the least glamorous, most paperwork-heavy corner of the economy: manufacturing and distribution.

There is a specific kind of misery familiar to anyone who works in construction, law, or manufacturing: you open a 6,000-page PDF, press Cmd-F, type a word, and start scrolling through hundreds of matches hoping one of them is the paragraph you actually need. That small, repeated frustration is where Ambria began - though back then it went by a different name.

Ambria is a San Francisco startup founded by Praneeth Guduguntla and Arman Rafati. It went through Y Combinator's Winter 2024 batch, and it now describes itself in four words: "AI for the Industrial world." The company builds AI agents aimed at the back office of manufacturing and distribution - the reading, checking, and processing of operational documents that keeps a factory or a distributor running, and that almost nobody enjoys doing.

To understand where Ambria is going, it helps to know where it started.

01 / THE FIRST PRODUCTA search engine for the documents nobody reads

In its YC batch, the company launched a product called Glimmer: an AI search engine for massive PDFs. You uploaded a document - even one running past 10,000 pages, well beyond what most AI tools would accept - and Glimmer chunked and indexed it. Then you asked questions in plain English and got answers back with verifiable page citations, so you could click straight to the source rather than trust a chatbot's summary.

The target users were people who live inside enormous documents every day: construction project managers buried in specifications, lawyers and paralegals combing contracts, finance teams, and healthcare coding and insurance specialists. The tool was free and self-serve. The pitch was blunt - the industry standard was still "Cmd-F keyword search and manually going through thousands of results," and Glimmer claimed to make that roughly ten times faster.

The best software is built with taste. Praneeth Guduguntla, Co-Founder & CEO

It was a clean, demonstrable idea, and it did what a good first product is supposed to do: it proved the team could reason reliably over documents too big for general-purpose AI to swallow. What it also revealed was where the real money and the real pain lived - not in consumer search, but in the operations of industrial companies.

02 / THE PIVOTFrom reading documents to running the work

Somewhere along the way, the company reframed itself as Ambria. The through-line is the technology - reasoning over dense, unglamorous documents - but the ambition grew. Instead of handing a professional a faster way to search, Ambria set out to build agents that do the back-office work itself: the routine processing that surrounds purchase orders, spec sheets, compliance paperwork, and distribution records.

Praneeth has described the current work plainly: building production-ready AI agents to automate back-office operations in manufacturing and distribution. It is a bet that the most valuable place for an agent is not writing marketing copy, but sitting inside an industrial company's paperwork and quietly getting through it.

The shift, in one line

Glimmer
Help a person search a giant PDF faster.
Ambria
Let an agent do the back-office work around it.
Same document-reasoning engine, aimed at a bigger, less crowded room.

This is a familiar move for anyone who has watched startups closely: the first product is a hypothesis, not a destination. The team kept the part that worked and pointed it somewhere with deeper budgets and fewer competitors chasing the same demo.

03 / THE FOUNDERSTen years of building, then Y Combinator

Praneeth Guduguntla and Arman Rafati are not first-time collaborators. They met as freshmen at the University of Illinois Urbana-Champaign and have been shipping software together, on and off, ever since - from nonprofit tools to viral side projects.

Their resumes read like two halves of a company. Praneeth, the CEO, built computer-vision tools at Palantir and worked on AI video editing for Facebook Reels at Meta. Arman, the CTO, worked on corporate cards at Mercury, helped build a first-of-its-kind Buy-Now-Pay-Later debit card at Affirm, and had a hand in launching AWS App Runner. One brings applied AI and product taste; the other brings hard infrastructure and fintech-grade reliability. Their YC partner was Pete Koomen, co-founder of Optimizely.

W24
Y Combinator batch
2
Co-founders, UIUC
10,000+
PDF pages Glimmer handled
~10×
Claimed search speedup

04 / THE MARKETWhy the industrial back office is the interesting bet

Manufacturing and distribution are enormous, and they run on documents. Every order, part, contract, and compliance requirement generates paperwork that someone has to read and act on. General assistants like ChatGPT and Microsoft Copilot are useful, but they were never built to understand a specific distributor's operations, and they tend to choke on the sheer size of industrial files. That gap - between general fluency and specific, verifiable, operational answers - is exactly the space Ambria is trying to occupy.

The competition is real but recognizable. On one side sit the ERP giants adding AI agents to their suites - Infor's Coleman, Acumatica, Epicor, and Microsoft Dynamics Copilot among them. On the other are other vertical AI startups circling back-office automation. A two-person company competing here is betting on focus and speed: going deeper on one workflow than a sprawling ERP roadmap can, and shipping with more taste than a generic assistant.

There is also a labor argument underneath all of this. Manufacturers and distributors have spent the last few years short-staffed and stretched, and much of the work that goes unfilled is exactly this kind of document-bound processing - not the skilled work on the floor, but the paperwork that surrounds it. An agent that can reliably read a spec, reconcile an order, or flag a mismatch is not competing with a craftsperson; it is competing with an overflowing inbox and a Friday-afternoon backlog. That framing - agents doing the reading so people can do the deciding - is the version of AI-at-work that industrial buyers tend to find easiest to say yes to.

The customers, then, are not consumers hunting a fact in a manual. They are the back-office teams at manufacturing and distribution companies, the people whose day is a queue of documents and exceptions. In Glimmer's earlier life those users were adjacent - construction managers, paralegals, healthcare coders - all of whom shared the same underlying problem of too much text and too little time. Ambria is following that thread to where the operational stakes, and the willingness to pay, are highest.

Where Ambria sits in the AI-agent landscape

Ambria
industrial-specific
ERP agents
bundled in suites
ChatGPT / Copilot
general-purpose
Illustrative positioning: the further right, the more specialized to industrial back-office work. Ambria's whole thesis is that specialization wins where operations are on the line.

05 / THE MODELWhat it is, and what is still unknown

Ambria is a B2B software company. Glimmer, the original product, was free and self-serve - a way to build a top of the funnel and prove the technology. The industrial-agents direction points toward the usual enterprise motion: subscriptions and contracts sized to the value of the work being automated. Exact pricing, customer names, and funding beyond the standard YC arrangement are not public, and the team remains small and early. Where the company lands on packaging is still being written.

That early-and-quiet status is worth stating plainly rather than dressing up. What is clear is the shape of the bet: a founding team with unusually broad engineering range, a proven ability to reason over documents that defeat other tools, and a decision to walk toward the industrial world rather than the crowded consumer-AI center.

AI for the Industrial world. Ambria company tagline

06 / THE TIMELINEHow Ambria got here

  • 2019
    Two builders meet at UIUC
    Praneeth and Arman start shipping software together as freshmen.
  • 2020–2023
    Big-tech tours
    Palantir and Meta for one; Affirm, Mercury, and AWS for the other.
  • 2024
    Y Combinator W24 & the Glimmer launch
    The pair reunite and launch AI search for massive PDFs on Launch YC.
  • 2024
    Re-positioning as Ambria
    The mission becomes "AI for the Industrial world," aimed at back-office automation.
  • 2025
    Talking industrial agents in public
    Praneeth presents on production-ready agents for manufacturing back offices.

At a glance

Company
Ambria (formerly Glimmer)
Batch
Y Combinator, Winter 2024
HQ
San Francisco, United States
Founders
Praneeth Guduguntla (CEO), Arman Rafati (CTO)
Focus
AI agents for manufacturing & distribution back offices
Team
2 (small, early-stage)

The through-line from Glimmer to Ambria is simple enough to fit on a whiteboard: the industrial economy is full of documents that nobody has the time to read carefully, and that is precisely the work an AI agent can take on. Whether Ambria becomes the company that owns that layer is an open question. But the instinct - follow the paperwork, and go where the general-purpose tools cannot - is a good one to watch.

#ai-agents#industrial-ai#manufacturing #distribution#back-office-automation#yc-w24 #pdf-search#vertical-ai#san-francisco