This startup fixes AI model drift in high-stakes workflows

This article is a part of Startup Spotlight, a series that features young, up-and-coming startups.

Image credit: Timmy Loen

In their years of deep learning research, Aakash Patil and Mrunmayee Shende noticed highly capable AI models consistently stalling in legal and financial workflows. The root cause was a fundamental lack of output reliability.

They realized the mathematical foundations for reliable AI were still unwritten. This gap led them to build Nugen Intelligence to provide more secure and reliable enterprise AI infrastructure.

😟 Problem

Generative AI struggles with sub-30% reliability in critical domains like legal contracts and technical specifications. In high-stakes workflows, an AI model that drifts once cannot be safely deployed at scale.

The average cost per critical enterprise AI failure has reached US$4 million. A recent Gartner study indicates 89% of CxOs are actively concerned about this reliability gap.

💡 Solution

Nugen Intelligence provides a low-code platform that aligns general-purpose AI models to specific domain data for critical sectors. The platform:

• Runs a live control loop to detect and correct behavioral drift during inference.
• Offers a low-code interface for uploading documents and deploying aligned AI agents.
• Applies selective layer adaptation for train-time alignment to specific domain data.
• Enables seamless migration for existing OpenAI users through a single-line codebase modification.

Image credit: Nudgen Intelligence

📊 Market size

Enterprises across the finance, healthcare, and legal sectors face urgent needs to mitigate regulatory penalties. Capturing a fraction of these commercial deployments addresses a multibillion-dollar infrastructure requirement.

🤝 Team

• Aakash Patil. Co-founder and CEO. Has a deep learning Ph.D., and is former Stanford machine learning scientist.
• Mrunmayee Shende. Co-founder and COO. Deeptech operator with experience scaling go-to-market strategies across Europe, the Middle East, and Africa, as well as Asia-Pacific and Japan.

🚀 Traction

• Built a seven-figure commercial pipeline involving more than 200 AI-native companies.
• Expanded platform capabilities from text into multimodal AI spanning vision and text.
• Selected to present research at the upcoming ICLR 2026 conference.
• Shortlisted for a US$3 million non-dilutive government research grant.

🏆 Competition

Standard fine-tuning locks AI behavior at training time and decays as business needs evolve. Nugen instead runs a live control loop to correct behavioral drift during active use. The platform relies on a compound AI approach that allows deployed models to self-improve over time.

💰 Financials

🚩 Risks