News
Obin AI Announces New Collaboration with Google Cloud
Obin Marketing
•

New York, NY, April 15, 2026 — Obin AI, an enterprise AI company building an agentic workforce for financial institutions, today announced its availability on Google Cloud Marketplace, enabling firms to rapidly deploy AI agents that expand operational capacity, sharpen risk decisions, and deepen client relationships.
Google Cloud Marketplace lets users quickly deploy functional software packages on Google Cloud, including Compute Engine and Cloud Storage, with no manual configuration required.
Through Obin AI’s availability on Google Cloud Marketplace, financial institutions can now more seamlessly access the Obin Agent Development Platform, which includes Obin’s Context Studio, Agent Harness, and Agent Accelerator toolkits. With these capabilities, firms can expand operational capacity, better tailor offerings for wealth and insurance customers, deploy capital more efficiently, and price risk more accurately.
Unlike platforms that require enterprises to hand over proprietary data to closed ecosystems, Obin AI is built on an open architecture. This approach ensures that institutions retain full ownership and control of their models, data, and intellectual property.
“Bringing the Obin Agent Development Platform to Google Cloud Marketplace will help customers quickly deploy, manage, and grow the financial agentic workforce on Google Cloud's trusted, global infrastructure," said Dai Vu, Managing Director, Marketplace & ISV GTM Programs at Google Cloud. “Obin AI can now securely scale and support customers on their digital transformation journeys.”
“By being available on Google Cloud Marketplace, Obin AI can better serve our shared Google Cloud customers. The tight integration of Gemini and Google Cloud Agent Development Kit (ADK) with Google Cloud’s strong data and AI infrastructure accelerates Obin’s development of agentic AI solutions in finance and makes our deployments robust.” — Lak Lakshmanan, CTO and Co-Founder of Obin AI
To mark this new collaboration, Obin AI’s CTO will present at Google Cloud Next on April 23, 2026, sharing real-world insights on building the granular traceability and observability workflows required by enterprise financial institutions using Google Cloud.
To learn more about how Obin AI can accelerate your core workflows and deploy the platform today, visit Obin AI on Google Cloud Marketplace or contact our business development team.
About Obin AI
Obin AI is an enterprise artificial intelligence company building the next generation of agentic intelligence for financial services. The company develops a production-ready AI workforce that runs defined workflows end-to-end inside a firm’s controls and audit boundaries, encoding institution-specific logic and producing traceable, inspectable outputs. Because of its open, no-lock-in architecture and ability to resolve even nuanced edge cases, Obin AI is trusted by global financial institutions representing over $1 trillion in assets under management. Founded by seasoned leaders from Google AI, JPMorgan and Silver Lake, Obin AI is reimagining how financial firms operate — deploying AI workers directly within live workflows to expand decision capacity under institutional controls.

News
Aug 25, 2026
Obin AI Brings Production-Grade Financial Agents to Google Cloud's Gemini Enterprise for Financial Services
The Obin Financial Agent is now available in Gemini Enterprise for Financial Services. Here is what it already runs in production at one of the largest U.S. life insurers and at a private credit fund with more than $500 billion in assets.

Engineering
Aug 20, 2026
Optimizing AI-Generated Content
Traditional A/B testing breaks down when you don’t know which characteristics actually make content perform better. Optimizing to the Eval solves this by generating competing outputs, choosing winners through human judgment, an evaluator, or real-world outcomes, and using those preferences to tune the model with DPO. The model never needs to learn why one style wins—it simply learns to produce more of what the evaluation rewards. The catch is that your evaluator becomes the product specification: if it reflects the outcome you actually care about, preference tuning can continuously improve quality; if it rewards the wrong proxy, the model will learn to game it.