hari User

At Innodata, I evaluate AI-generated responses to improve the factual accuracy, reasoning quality, and linguistic consistency of Large Language Models. I maintain 95%+ quality scores across high-volume AI evaluation tasks.

I perform RLHF, prompt evaluation, response ranking, and content validation to improve model safety, alignment, and instruction-following. I also annotate and validate text, image, and audio datasets, reviewing complex edge cases against strict project guidelines.

Previously at PwC, I analyzed financial statements, annual reports, and business performance for multinational clients. I prepared transfer pricing documentation, supported APA and MAP engagements, and delivered research-based recommendations while maintaining regulatory compliance.

My foundation in accounting, auditing, taxation, and financial analysis comes from audit work at Ramalingam & Associates and operations work in real estate. I bring detailed analytical reasoning, documentation discipline, and finance-domain context to AI quality work.