Nischal Nadhamuni

Nischal Nadhamuni is co-founder and chief technology officer of Within, formerly Klarity, an enterprise AI company building a Company Brain to capture organizational knowledge and coordinate work between employees and AI agents. His products have progressed from automated contract review to finance workflows and enterprise-wide operational intelligence.

Nadhamuni studied computer science at MIT, graduating in 2018, and worked on AI projects at Flipkart, Massachusetts General Hospital, and Airware. At Flipkart, he developed a system for detecting fraudulent user behavior. He met future co-founder Andrew Antos in an MIT entrepreneurship class in 2016; they founded Klarity in 2017 and joined Y Combinator’s summer 2018 cohort.

Their initial product analyzed contracts against customer requirements. After several pivots, Klarity concentrated on finance and accounting, automating revenue recognition, invoice matching, and multilingual tax-document processing. The company raised $18 million in 2022 and announced a $70 million Series B in 2024.

Nadhamuni subsequently rebuilt Klarity’s platform around generative AI, establishing a dedicated engineering team, security safeguards, and red-team testing during a roughly ten-week transition. Under the Within name, the company expanded beyond finance into institutional knowledge, workflows, decisions, and collaboration between people and agents.

• Customer-specific AI evaluation: Measure the business outcome first, then adoption, feature health, and task accuracy. Nadhamuni favors customer-labeled documents, acceptance testing, feedback, and document-drift monitoring because public model benchmarks frequently fail to predict performance on actual finance workflows. His account of enterprise evaluation also describes synthetic test data designed to resemble production documents.

• Validate new product experiences early: Features such as natural-language analytics and automatically generated business-requirements documents introduce unfamiliar usability risks. Nadhamuni tests customer demand before building extensive evaluation infrastructure, while requiring stronger safeguards before production deployment.

• Keep automated prompt engineering tractable: Customer-specific prompts multiplied across features and models can create thousands of combinations. Standardizing on fewer model choices can make systems easier to evaluate and operate, even when exhaustive optimization might yield marginal accuracy gains.

• Enterprise context graphs must capture human judgment: Databases preserve transactions more readily than exceptions, workarounds, and precedents.