Deep Cogito Raises $43M Series A to Advance the Post-Training Engine for Frontier Intelligence

Key Terms

post-training technical

The period or set of steps that occur after a machine-learning model has completed its initial training, including evaluation, validation, fine‑tuning, calibration, optimization for speed or size, and preparation for deployment or regulatory review. It matters to investors because these activities determine whether a model works as intended in the real world, affect time‑to‑market, compliance risk, ongoing costs, and the practical value of any product or service that relies on the model—like tuning a prototype into a reliable tool.

reinforcement learning technical

A type of artificial intelligence that learns by trial and error, receiving feedback from its actions to favor choices that lead to better outcomes. Think of it like a salesperson learning which pitches close deals by trying different approaches and keeping the ones that work. For investors, reinforcement learning matters because it can power smarter trading systems, optimize business operations, or improve products—potentially boosting efficiency and profits while also introducing model and execution risks.

recursive self-improvement technical

A process in which a software system or artificial intelligence makes changes to its own code, model or design in order to improve its performance or capabilities without direct human reprogramming. It matters to investors because it can rapidly accelerate a company's product development, competitive edge and potential value—like a machine that upgrades itself—while also changing the speed and uncertainty of technical progress and the regulatory and safety risks a business faces.

iterated distillation and amplification technical

A machine‑learning training method that builds more capable AI by repeatedly having a set of systems or processes collaborate to solve tasks (amplification) and then compressing their combined behavior into a single, smaller model (distillation). Think of many tutors working together to solve problems and then producing a concise textbook that teaches a single student. It matters to investors because it can change how quickly firms improve AI performance, scale products, and address safety or reliability concerns.

open-weight models technical

Open-weight models are machine-learning systems whose internal parameters (the numeric “weights” that determine how inputs are turned into outputs) are publicly available for inspection, modification, and reuse. For investors, that transparency makes it easier to audit performance claims, assess risks and biases, and accelerate product development—think of it as a recipe where every ingredient and measurement is shown, so anyone can test, tweak, or reproduce the result.

See more from StockTitan in Google Search and AI answers. Adds StockTitan as a preferred source · opens Google

Add on Google Not now

• Founded by the team behind Google AI Search, Deep Cogito is building large-scale reinforcement learning and self-improvement systems for frontier AI.
• Its post-training engine powers both the Cogito family of open-weight models and specialized models trained on enterprises' proprietary data and outcomes.
• TQ Ventures led the round, with participation from Benchmark, Nexus Venture Partners, Atreides Management, South Park Commons, and customer and strategic investor Zscaler.

SAN FRANCISCO--(BUSINESS WIRE)-- Deep Cogito, a post-training research lab focused on reinforcement learning and self-improvement, today announced a $43 million Series A led by TQ Ventures, with participation from Benchmark, Nexus Venture Partners, Atreides Management, South Park Commons, and Zscaler. The round brings Deep Cogito's total funding to more than $56 million.

Deep Cogito was founded by Drishan Arora and Dhruv Malrana, who previously helped build Google's AI Search products, including AI Mode and AI Overviews. At Google, Arora led Gemini post-training for AI Search, while Malrana led its product from inception.

They started the company around a single thesis: as AI advances, most of the frontier will be determined by post-training, the process that turns a pre-trained model into a capable reasoner and teaches it to improve on increasingly difficult tasks.

The company's research focuses on large-scale reinforcement learning and recursive self-improvement. One of its research directions, Iterated Distillation and Amplification (IDA), repeatedly allows a model to use additional computation to produce answers beyond what it could generate directly, then distills those improvements back into the model's weights. The long-term goal is to build models that progressively improve their own capabilities and ultimately move beyond the limits of human-generated training data.

"Pre-training gives a model an enormous amount of knowledge and capability. Post-training determines what that model can actually become," said Drishan Arora, co-founder and CEO of Deep Cogito. "We believe the next frontier is in finding ways for models to improve their own intelligence, internalize those improvements, and become increasingly capable over time."

Deep Cogito first developed its post-training methods through its open-weight model releases. Across model sizes from 3B to 600B+, that work showed the company could improve strong models through post-training and reinforcement learning. The same system now powers the platform Deep Cogito is making available to companies that want to build specialized intelligence for their own products.

"Very few teams outside the largest AI labs have demonstrated the ability to post-train models at this scale," said Schuster Tanger, Co-Founding Partner at TQ Ventures. "Deep Cogito has done that in public through its model releases, and is now bringing the same capability to companies that want intelligence built around their own products. We believe that combination of frontier research and real-world deployment is extremely powerful."

Zscaler (NASDAQ: ZS), a leader in cloud security, began working with Deep Cogito as a customer and is also participating in the Series A as a strategic investor.

"Frontier models were useful, but they were not enough for the level of specialization we needed," said Dhawal Sharma, Executive Vice President of AI Security and Strategic Initiatives at Zscaler. "Deep Cogito stood out because they went deeper than lightweight customization. They worked closely with us to understand our products and the metrics we care about and helped train that intelligence into the model itself."

Deep Cogito will use the new capital to expand its research and engineering team, scale the infrastructure required to train frontier models, advance future Cogito releases, and grow its work with enterprises building specialized intelligence on proprietary data.

"What stood out to us was not only the technical depth of the team, but the scope of what they're trying to build," said Eric Vishria, General Partner at Benchmark. "Post-training is becoming one of the most important layers in AI. Deep Cogito has demonstrated that it can operate at the frontier of that layer and translate that capability into intelligence that companies can actually own."

For more information, visit deepcogito.com.

About Deep Cogito

Deep Cogito is a post-training research lab focused on reinforcement learning and recursive self-improvement. The company builds frontier open-weight models through the Cogito family and applies the same post-training engine to create specialized models trained on enterprises' proprietary data, decisions and outcomes. Deep Cogito was founded by members of the team behind Google AI Search and is based in San Francisco.

View source version on businesswire.com: https://www.businesswire.com/news/home/20260826913379/en/

VSC on behalf of Deep Cogito
deepcogito@vsc.co

Source: Deep Cogito