Deep Cogito has raised $43 million in Series A funding to expand its post-training research platform for reinforcement learning, recursive self-improvement and enterprise AI specialization.
TQ Ventures led the round, with participation from Benchmark, Nexus Venture Partners, Atreides Management, South Park Commons and Zscaler.
The financing brings Deep Cogito’s total funding to more than $56 million.
Deep Cogito was founded by Drishan Arora and Dhruv Malrana, who previously worked on Google’s AI Search products, including AI Mode and AI Overviews.
At Google, Arora led Gemini post-training for AI Search, while Malrana led product development from inception.
The company is built around the thesis that post-training will become increasingly important as frontier AI advances.
Pre-training gives models broad knowledge and capabilities, while post-training is used to shape how those models reason, adapt and improve on increasingly difficult tasks.
Deep Cogito’s research focuses heavily on large-scale reinforcement learning and recursive self-improvement.
One of its core research directions is Iterated Distillation and Amplification, or IDA.
The approach allows a model to use additional computation to produce stronger answers than it could generate directly, then distills those improvements back into the model’s weights.
Over time, Deep Cogito aims to create models that can progressively internalize improvements and move beyond the limitations of human-generated training data.
The company first developed and tested its post-training methods through its Cogito family of open-weight models.
Those releases span model sizes ranging from 3 billion parameters to more than 600 billion parameters.
Deep Cogito said the work demonstrated its ability to improve already capable models through reinforcement learning and post-training at large scale.
The same technology now underpins an enterprise platform designed to help companies build specialized AI models using their own proprietary data, decisions and outcomes.
Rather than relying primarily on lightweight customization, Deep Cogito works with enterprises to train domain-specific capabilities directly into models.
Zscaler is both a customer and a strategic investor in the Series A.
The cybersecurity company worked with Deep Cogito to train specialized intelligence around its products and internal performance metrics.
Deep Cogito plans to use the new funding to expand its research and engineering teams, increase the infrastructure available for frontier-model training and advance future Cogito model releases.
The company will also expand its work with enterprises seeking to build proprietary AI intelligence around their own data and business outcomes.
KEY QUOTES:
“Pre-training gives a model an enormous amount of knowledge and capability. Post-training determines what that model can actually become. 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.”
Drishan Arora, Co-Founder And CEO Of Deep Cogito
“Very few teams outside the largest AI labs have demonstrated the ability to post-train models at this scale. 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.”
Schuster Tanger, Co-Founding Partner At TQ Ventures
“Frontier models were useful, but they were not enough for the level of specialization we needed. 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.”
Dhawal Sharma, Executive Vice President Of AI Security And Strategic Initiatives At Zscaler
“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.”
Eric Vishria, General Partner At Benchmark
