Who is Adarsh Agrahari?

Adarsh Agrahari is a product leader working on AI products and platforms. He is Associate Director, Product & Design at HROne, where he led One AI, a multi-agent HR copilot. Earlier he built the Spend Power credit decisioning engine at Paytail, the Work Graph at Soroco, and SolarPulse predictive maintenance at Mahindra.

What has Adarsh Agrahari built?

One AI at HROne, a multi-agent HR copilot running in production across enterprise orgs. Spend Power at Paytail, a credit decisioning engine that scaled from pilot to a pan-India lending book. SolarPulse at Mahindra, anomaly detection across a gigawatt-scale renewable portfolio. The Work Graph at Soroco, which models how enterprise teams actually work.

What companies has Adarsh Agrahari worked at?

HROne (Associate Director, Product & Design), Paytail (Senior Product Manager), Soroco (Product Manager, AI & Automation), and Mahindra Group (Product Manager, Renewable Energy). He has also advised Doubtnut, Quarical.ai, Bitgrit, and Relevel by Unacademy, and founded HCS Business Solutions and GSpin.

Where does Adarsh Agrahari teach?

He teaches GenAI Product Development as visiting faculty at Masters' Union, product management at ISB Executive Education, and has taught at Mesa School of Business and Upraised. He has spoken at XLRI Jamshedpur and ProductTank.

How does Adarsh Agrahari approach multi-agent AI systems?

Specialized agents for each domain rather than one model doing everything, a lightweight orchestrator to route between them, vector memory for context retention, and human-on-loop guardrails wherever a decision has consequences. One AI at HROne uses this architecture across payroll, leave, and performance.

What does Adarsh Agrahari write about?

Essays on product management, AI governance, and systems thinking, plus case-study teardowns of the systems he has built. He has also published books on personal finance and product development.

LLM Reasoning Patterns | Adarsh Agrahari

< img src = "https://images.unsplash.com/photo-1620712943543-bcc4688e7485?q=80&w=1965&auto=format&fit=crop" alt = "LLM Reasoning" class="w-full rounded-lg mb-8 border border-white/10" />

Understanding how Large Language Models arrive at conclusions is critical for debugging agentic behaviors.

< h3 > Comparison < p > Tree of Thoughts(ToT): Explores multiple reasoning paths, backtracking when necessary. Ideal for complex problem solving.

Chain of Density(CoD): Iteratively compresses information to increase density without losing context. Great for summarization.

LLM Reasoning Patterns

More from Mind Map

The Agentic Workflow