EO StudioThis Is How a 26-Year-Old Raised $108M in 1.5 Years | Reducto, Adit Abraham
CHAPTERS
- 0:00 – 0:31
Doing the unsexy manual work to get the first customers over the line
Adit describes how Reducto’s early progress came from hands-on, repetitive work—manual labeling, manual billing setup, and whatever else was needed to keep customers moving forward. He frames this as a mindset: the goal isn’t glamorous work, it’s momentum and customer success.
- •Early-stage progress required heavy manual effort (labeling, QA, onboarding, billing)
- •Founders directly did the work when early hires/tools weren’t sufficient
- •Seeing repetitive tasks as a privilege because they indicate real customer demand
- •Focus on forward motion: “Is the company moving forward?”
- 0:31 – 1:31
What Reducto does and how it scaled fast (customers, pages, funding)
He introduces Reducto as an unstructured-data parsing and extraction platform for AI teams. Adit highlights rapid growth: major enterprise customers, leading AI startups using the product, and $108M raised while processing over a billion pages.
- •Reducto helps parse/extract/edit complex unstructured data for LLM use cases
- •Backed by top investors; raised $108M in ~1.5 years (per video framing)
- •Used by Fortune 10 enterprises and AI-native companies
- •Scale metrics: >1B pages processed and growing weekly
- 1:31 – 2:01
Early hacker mindset: side projects, apps, and outlier ambition
Adit traces his entrepreneurial drive back to high school, inspired by stories like Flappy Bird’s success. The takeaway is the power of side projects and unconventional paths as seeds for larger outcomes.
- •Motivation sparked by seeing small projects become huge successes
- •Early experimentation: trying to build apps rather than follow a standard path
- •Lessons from “off the beaten path” thinking
- •Side-project energy as a precursor to founding a company
- 2:01 – 3:02
Meeting Ronak at MIT and choosing a cofounder you’d bet on immediately
At MIT, Adit meets Ronak in an advanced ML course where Ronak stands out immediately. Adit explains why working with someone exceptional made the decision to build together feel obvious and low-friction.
- •MIT experience and exposure to advanced ML early on
- •Ronak’s standout capability signaled rare talent and leadership
- •Cofounder choice based on admiration/trust and complementary strength
- •Deciding quickly when the right partner appears
- 3:02 – 3:33
Killing ‘nice-to-have’ ideas: chasing urgent pull instead of future need
Before YC, they tested multiple products and even walked away from early revenue because customer urgency wasn’t strong enough. Adit contrasts a product people might need someday with one they need immediately—and how that shaped their pivot decisions.
- •They abandoned revenue-generating ideas when urgency was low
- •Remember All: a long-term memory API for LLMs, priced modestly ($50–$100/mo)
- •Customer feedback revealed it wasn’t the top pain point or priority
- •Strategic filter: build for present pull, not speculative future demand
- 3:33 – 5:03
The pivot moment: file/document handling became the product customers demanded
A “feature request” around managing uploaded files turned into the most compelling value. When customers reacted with immediate buying signals—asking for an API, Stripe link, and switching from vendors—it revealed strong product-market pull.
- •File management for uploads started as an add-on to Remember All
- •They improved document segmentation and built a simple demo (Streamlit + box drawing)
- •Users compared results to existing vendors and wanted to buy immediately
- •Clear market signal: inbound purchasing intent and stronger urgency than the original product
- 5:03 – 6:34
Building with customers as design partners: production edge cases and fast iteration
Adit explains that real production documents (finance, healthcare, insurance) contain edge cases you won’t find online. Reducto’s advantage came from tight customer feedback loops—iterating daily on hard examples to improve models quickly.
- •Production data differs dramatically from public benchmarks/documents
- •Customers bring extremely hard edge cases (annotations, complex tables, unusual layouts)
- •Design-partner workflow: rapid feedback → model iteration → validation
- •Customer collaboration becomes a moat for accuracy and robustness
- 6:34 – 7:34
High-touch customer success as a differentiator: Slack channels, phone numbers, paging founders
From the beginning, Reducto operated like an extension of the customer’s ingestion team. They used direct communication channels and urgent support to repay early trust and ensure mission-critical workflows didn’t break.
- •Dedicated Slack channels per customer; founders directly reachable
- •Fast response culture for customer issues and escalations
- •Trust-building: earning credibility vs. established incumbents
- •Reducto sold not only software, but an embedded ingestion team experience
- 7:34 – 8:34
Don’t just claim accuracy—prove it publicly with a playground and hard-document demos
Rather than relying on marketing claims, Reducto put the product in front of users—even before the platform was perfect. A public playground let skeptical prospects test their hardest documents, turning proof into pipeline (including major enterprises).
- •Differentiation through demonstrable results vs. generic “state-of-the-art” claims
- •Public playground enabled self-serve validation on difficult documents
- •Credibility booster: even a two-person team could attract trillion-dollar enterprises
- •“Show, don’t explain” created faster sales cycles and more inbound
- 8:34 – 9:08
Compounding trust: prospects who said ‘not yet’ came back after seeing steady progress
Adit notes that some companies hesitated to buy from an early-stage vendor even if they liked the product. Continued visible improvement month-over-month converted many of those initial ‘no’s into later inbound ‘yes’s.
- •Early-stage risk perception slowed some deals despite product quality
- •Consistent execution and visible progress reduced buyer fear over time
- •Follow-on inbound from previously skeptical prospects
- •Year-one sales resistance turning into year-two customers
- 9:08 – 10:08
Choosing investors for the hard times: evaluating the partner, not the firm brand
He argues founders should optimize for the individual partner relationship over firm prestige. The key test is how an investor behaves when things go wrong—because they’ll be in the loop for a decade of both wins and setbacks.
- •Founder mistake: overweighting VC brand vs. partner fit
- •Investors are long-term partners across good and bad moments
- •Assess how communication/support changes under stress
- •Fundraising as one of the most consequential long-term decisions
- 10:08 – 11:09
What supportive investors look like in practice: the outage story
Adit shares a concrete example of their seed investor Liz stepping in during an OpenAI outage, even while off-duty. She leveraged her network to rapidly reach senior help—demonstrating commitment beyond typical expectations.
- •Real-time crisis support is a critical investor attribute
- •Investor responsiveness and proactive problem-solving during outages
- •Network access can materially reduce downtime and customer impact
- •Signals of commitment: showing up when it’s inconvenient
- 11:09 – 12:10
The unseen grind behind big enterprise wins: on-prem deployments and all-hands execution
Landing major enterprise contracts required intense execution, including an on-prem deployment they’d never done. Adit emphasizes an all-hands culture where everyone does whatever is necessary—support, ops, labeling—because outcomes matter more than titles.
- •Enterprise deal requirements forced new capabilities (on-prem) under pressure
- •Founders worked extreme hours to deliver with a small team
- •No role silos: everyone contributes where needed for customer success
- •Culture defined by ownership, urgency, and making customers happy
- 12:10 – 13:00
Reducto’s broader vision: the context layer powering agentic workflows
Adit frames Reducto as more than parsing—it's a bridge between human data and AI intelligence. He envisions Reducto as a core building block for context that enables end-to-end, agentic products that can read, understand, and generate new documents.
- •Reducto as a “context layer” connecting messy data to AI intelligence
- •Customers building agentic workflows that generate net-new documents
- •Future AI products = intelligence (models) + context (data infrastructure)
- •Goal: be the best building block for interacting with and applying context
