Harsh Sonbhadra

About Harsh Sonbhadra

I am an AIBE-qualified Advocate with hands-on experience in civil and criminal litigation, legal drafting, and courtroom procedure, developed through practice at a Delhi-based law firm. I have worked on AI training and evaluation projects across generalist, language, and legal domains, contributing to the development of safe, reliable, and high-quality AI systems. My work included voice and audio annotation, legal and commonsense reasoning tasks, metadata tagging, rubric-based evaluations, and red-teaming of model outputs, with a strong focus on quality assurance, consistency, and safety. Alongside applied AI work, I have researched and written on cyber law, data privacy, intermediary liability, and child online safety. This work strengthened my understanding of regulatory thinking, compliance, and digital governance in technology-driven environments. I am building a career at the intersection of law and artificial intelligence, with a growing focus on data protection, privacy, AI governance, trust and safety, and responsible AI systems. My goal is to work in roles where legal reasoning and applied AI operations come together to support ethical, compliant, and human-centric technology.

Experience

1. Ai Training Generalist

   Scale AI

   Nov 2025 — Present · San Francisco, CA, US

   Currently working on large-scale AI training and evaluation tasks focused on improving model behavior, naturalness, and reliability. Design and review high-quality prompts, evaluate model responses against detailed rubrics, and identify failure modes under defined constraints. Contribute to dataset diversity, consistency checks, and quality assurance across text, audio, and multimodal tasks while adhering to strict accuracy and guideline standards.

Education

• Kendriya Vidyalaya

  10+2, Non-Medical with Computer Science

  2016 — 2018

• Vivekananda Institute of Professional Studies

  Bachelor of Laws - LLB, Law

  2018 — 2023