Founding AI Engineer at Genloop - building the hard parts of making data answer for itself.

Kaushal Kumar • Peerlist

B2B agents work in the demo. They break when they meet a real enterprise.On sample data, agents look capable. At a real customer, they encounter that enterprise’s own tables, documents, policies, and people. They understand the product but lack the business context to answer reliably. Accuracy drops, trust erodes, and vendors send engineers to close the gaps—customer by customer.Genloop provides the context infrastructure for B2B agents. A software company onboards its product once. Genloop connects that product knowledge with each enterprise’s data, documents, rules, and people into one unified context that stays current and ships inside the product.•&nbsp; Living Context Graph: Maps relationships across product knowledge and enterprise data so agents understand how each customer’s business works.•&nbsp; Self-Learning Loop: Improves shared context through expert-approved corrections, feedback, and verified updates.•&nbsp; Grounded answers: Gives agents answers with citations to the underlying tables and documents.•&nbsp; Embedded infrastructure: Serves context through MCP and APIs, bringing enterprise understanding directly into your product.•&nbsp; Enterprise control: Keeps each customer isolated, with private deployment and governance over what the system learns.On a Fortune 500 company’s real data, Genloop improved the same frontier model from 30% to 94% correct answers. Initial onboarding took two weeks, compared with an estimated six months, followed by approximately three days per enterprise.Building the foundation on which SaaS thrives in the agent era.

Genloop

• Implemented a load balancing policy for a private cloud environment. • Analyzed the resource consumption of virtual machines to identify potential load balancing conflicts. • Developed a self adaptive weighted approach to resolve load balancing conflicts caused by different resource demands of cloud applications. • Incorporated a probabilistic approach to ease the problem of load crowding in the concurrent user scenario, and adjust the system accordingly achieving a significant jump in performance.