Get Your Hands Dirty Even in the Early Stage | Reducto, Adit Abraham

Reductto: From Manual Labeling to AI Data Ingestion – A Detailed Account

Key Concepts:

• Data Labeling: The process of identifying and annotating data to train machine learning models.
• Long-Term Memory API: An API allowing language models to retain and recall information from past interactions.
• Unstructured Data: Data that does not have a predefined format (e.g., documents, text files).
• Ingestion: The process of bringing data into a system for processing.
• Meta-learning: A machine learning approach where models learn how to learn.
• Agentic Workflows: AI systems capable of autonomous action and decision-making.
• Production Data: Real-world data used in operational systems, often differing from publicly available datasets.

I. Early Days & The Value of "Unsexy" Work

Ad Abraham, co-founder and CEO of Reductto, recounts the company’s humble beginnings, characterized by extensive manual labor. Initially, the team struggled with the accuracy of outsourced data labeling, leading Abraham to personally label thousands of document pages – a volume equivalent to ten times the height of Mount Everest. He also manually set up every Stripe subscription in the early days. Despite the repetitive and seemingly unglamorous nature of this work, Abraham emphasizes its importance: “The thing that I cared about is not like am I doing the most glamorous work. It was more so like is the company moving forward?” He frames this period as a “privilege” because each task signified acquiring a new customer. This highlights a core principle: prioritizing company progress over individual task prestige.

II. From Side Projects to MIT & The Genesis of Reductto

Abraham’s entrepreneurial spirit dates back to high school, inspired by the success of Flappy Bird and its $50,000 daily revenue. This early ambition, though initially unrealized, instilled a belief in the potential of side projects to achieve “outlier outcomes.” He eventually pursued computer science at MIT, where he encountered Ronic, a remarkably capable freshman who assisted in a graduate-level meta-learning course. Abraham was immediately impressed by Ronic’s aptitude and readily agreed to collaborate, stating, “The first time Ronic suggested that we could work on something together, that was an immediate yes for me.” This partnership proved foundational to Reductto’s development.

III. The Iterative Journey with "Remember All" & Pivoting to Data Ingestion

The initial company, preceding Reductto’s current form, focused on “Remember All,” a long-term memory API for language models. Despite achieving some initial traction with customers willing to pay $50-$100/month, the team found the urgency of need insufficient. Customers viewed long-term memory as a “nice to have” rather than a critical requirement.

A turning point occurred when they added file management capabilities to Remember All, initially as a simple feature built with off-the-shelf tools. Unexpectedly, customers became significantly more excited about this functionality. A simple Streamlit app demonstrating document segmentation, showcasing improved results compared to existing vendors, generated immediate purchase inquiries: “They immediately started replying with, 'Hey, these are better results than what I'm seeing from my existing vendor. Is this a hosted API? Do you have a Stripe link?'”

This realization led to a pivotal decision: to prioritize solving immediate customer needs over pursuing long-term, but currently less pressing, solutions. They concluded that Remember All, while valuable, wasn’t the problem customers were actively seeking to solve now.

IV. The Power of Production Data & Customer Collaboration

Reductto’s current success is rooted in its focus on processing complex, real-world data – “production data” – that is unavailable in public datasets. They serve intensive use cases in finance, healthcare, and insurance, handling challenging examples like doctor annotations and complex financial tables.

Abraham emphasizes a “design partner” relationship with customers, actively soliciting feedback and iterating on their models. This collaborative approach is crucial: “Our customers want us to solve those and so we've always had this almost design partner like relationship where they will come to us with that sort of feedback and we will iterate day after day after day to make the models better.” This iterative process, fueled by direct customer input, drives continuous improvement.

The company fosters close relationships with customers, maintaining individual Slack channels and direct phone access, demonstrating a commitment to providing personalized support: “even today you know if a company has an issue they can just pay Ronic or me directly.”

V. Growth Strategy: Public Demonstrations & Proving Value

Reductto’s growth strategy deviates from traditional marketing approaches. Instead of solely relying on claims of superiority, they prioritize demonstrating their product’s capabilities directly to potential customers. They made a public playground available, allowing users to upload their most challenging documents and experience Reductto’s performance firsthand.

This strategy proved remarkably effective: “At least in our case, being public in that way just meant that companies that otherwise probably would have ignored Reducto became really interested.” A trillion-dollar enterprise booked a demo specifically because of the public playground, validating Reductto’s claims. Abraham states, “if we hadn't done that, if we were this twoperson company of, you know, 20some year olds, I find it hard to imagine that they would even be interested in engaging with us.”

VI. Funding & The Importance of Investor Partnership

Abraham discusses the importance of choosing investors based on individual partner compatibility rather than solely on firm brand. He stresses the need for a long-term partnership, encompassing both successes and challenges: “They're going to be there in all of your great successes like your future fundraising rounds when you close the great contracts but they'll also be there for the bad moments of the company.” He cites an example of investor Liz’s unwavering support, even interrupting a personal event to assist during a critical outage, highlighting the commitment of a true partner.

VII. The Reductto Vision: Connecting Data to Intelligence

Abraham concludes by articulating Reductto’s broader vision: “what I see reductive as it, it's not really just parsing. It's what does it mean to have this layer that connects human data to this new level of intelligence that applies across all of that data.” He envisions a future where AI products combine foundation model intelligence with contextual understanding, and Reductto will be the leading platform for interacting with that context – a fundamental building block for AI applications. He predicts that future AI products will be a combination of intelligence and context, with Reductto positioned as the premier solution for managing and applying that context.

Data & Statistics:

• Funding: $108 million total funding.
• Pages Processed: Over 1 billion pages processed to date.
• Early Revenue (Remember All): $50-$100/month potential revenue.
• Mount Everest Analogy: Manual data labeling volume equivalent to 10x the height of Mount Everest.