Dhruva Myakeri

An agentic AI system that predicts and explains the ROI of a proposed AI project before a company spends the money. Not a calculator, not a static questionnaire: a multi-agent pipeline that reasons through the decision like a financial analyst would. Who it's for: a mid-size enterprise with a real, running operation (actual costs, actual volume data) deciding whether to adopt one specific AI capability. The problem: companies pour money into AI but can rarely answer what they got back, or will get back. Only ~20% achieve the AI revenue growth they expect, and unclear ROI is a top reason AI initiatives get killed. How it works: a Planner classifies the project and branches into scenarios when key details are unknown. A retrieval layer checks every company claim against cited industry benchmarks, flagging what doesn't hold up. A deterministic Python tool computes the ROI, no LLM does the math. An explainability agent scores confidence per dimension, and a report agent delivers a grounded recommendation. Built on Vultr Serverless Inference and NVIDIA Nemotron.