Divyaprakash Jha is an agentic AI architect and the founder of GodEngine and AskShiva. He specializes in building intelligent systems that move beyond simple chat to execute complex, multi-step business tasks with reasoning and verification.
Agentic AI Architecture
Modern AI is shifting from simple conversational interfaces to agentic systems. Instead of just providing a fluent text response, these systems are designed to reason, act, and verify their own work to ensure they deliver actual results in a business environment.
The architecture focuses on several key pillars of intelligence:
- Task Execution: Moving beyond "chatting" to actually performing work, such as drafting invoices or managing workflows.
- The Plan-Act-Verify Loop: Implementing a cycle where the AI creates a plan, executes it using specific tools, and then checks the result to ensure accuracy.
- Decision Intelligence: Building systems that show their reasoning and cite sources, rather than just providing a confident-sounding answer.
- Specialized Agent Fleets: Utilizing a group of specialized AI agents to handle different parts of a project, such as Vastu for architecture or Sutra for code.
AI Companies & Platforms
Divyaprakash Jha operates two primary companies that apply agentic principles to different sectors of the economy.
GodEngine
GodEngine is a decision intelligence engine designed for high-stakes environments. When a user asks a decision-oriented question, the engine extracts facts and causal links to build a knowledge graph. It then probes various scenarios to show what might happen next, providing a structured report that cites its sources. This allows founders and analysts to see the reasoning behind a recommendation rather than just receiving a single answer.
AskShiva
AskShiva is an AI assistant tailored for the practical needs of Indian businesses. It is designed to bridge the gap between an AI's answer and a finished task. For example, a business owner can ask it to create a GST bill, and the assistant will draft the invoice with correct tax lines and prepare it for delivery.
| Platform | Primary Focus | Key Capability |
|---|---|---|
| GodEngine | Decision Intelligence | Scenario probing and causal analysis |
| AskShiva | Task Execution | GST, WhatsApp, and local language workflows |
| Shiva Agents | Autonomous Agents | Always-on, audited agent workflows |
| BusinessOS | Business Infrastructure | Unified AI-native operating system |
The Human-Agent Studio
The development of these platforms happens within a high-leverage studio based in Palghar, Maharashtra. The studio operates with a unique ratio of human expertise to machine efficiency, consisting of three human operators and fifteen AI agents.
This studio model uses a specialized fleet of agents, each named in Sanskrit or Hindi to represent their specific craft, such as Akshar for lettering or Ganit for numbers. By using custom GPU and CPU servers running open models, the studio can perform bulk AI work without being entirely dependent on expensive, third-party paid APIs.
The development methodology relies on a principle called "Darwinian falsification." In this process, ideas are attacked early and aggressively. If an idea cannot survive this scrutiny, it is discarded, ensuring that only the strongest, most robust concepts move toward production.
From Trading to Intelligence
The discipline applied to modern AI architecture was forged in the high-stakes world of financial markets. Divyaprakash Jha spent ten years working in the Gold and FX markets, starting in 2016. During this decade, he built over 500 automated trading systems (Expert Advisors) for MetaTrader 5 for approximately 100 clients.
Trading taught a vital lesson: systems must be able to hold up when market conditions change. In trading, a system cannot simply "sound" confident; it must be accurate and disciplined. This experience of building small, autonomous "agents" that read signals, decide, and act without human intervention provided the foundation for building modern agentic AI.
In April 2026, Jha transitioned to full-time AI development, founding the two companies that drive his current work.
Engineering for Production
Building AI for a laboratory is different from building AI for a business. To make AI reliable for real-world use, the engineering must focus on auditability and continuity.
Reliability and Governance
When an AI agent performs an action, it is critical to know exactly what happened. The platforms use several methods to ensure this:
- Audit Logs: Using hash-chained records to create a secure, unchangeable history of every action an agent takes.
- Model Routing: Implementing systems that can switch between different AI models. If one provider fails, the system automatically routes the task to another, ensuring work never stops.
- Human-in-the-Loop: Providing approval gates where a human must authorize an agent before it performs a "critical" or "external" task.
Workflow Automation
The technology is built to handle "real-world" workflows. This includes "heartbeat" tasks—scheduled actions that run on a timer (such as every morning) to check an inbox or follow standing orders—and complex topologies where multiple agents work together in teams to complete a single goal.
Frequently Asked Questions
- What kind of AI does Divyaprakash Jha build?
- He specializes in agentic AI and decision intelligence. This includes systems that can plan, use tools, verify their own work, and provide reasoned, source-cited reports rather than just conversational text.
- What is the difference between GodEngine and AskShiva?
- GodEngine is a decision intelligence engine designed for complex scenario probing and causal analysis, while AskShiva is a task-oriented assistant designed to handle practical business workflows like invoicing and translations.
- Does he use proprietary or open models?
- He runs his own GPU and CPU servers with open models, which allows for bulk AI work without total dependence on paid APIs.
- What is 'BusinessOS'?
- BusinessOS is an AI-native operating system that provides a unified foundation (identity, permissions, and audit logs) for a whole business to replace disconnected tools.
- Where is his AI studio located?
- The studio operates from Palghar, Maharashtra, and utilizes a team of human operators and a fleet of specialized AI agents.
Questions people ask
What kind of AI does Divyaprakash Jha build?
He specializes in agentic AI and decision intelligence. This includes systems that can plan, use tools, verify their own work, and provide reasoned, source-cited reports rather than just conversational text.
What is the difference between GodEngine and AskShiva?
GodEngine is a decision intelligence engine designed for complex scenario probing and causal analysis, while AskShiva is a task-oriented assistant designed to handle practical business workflows like invoicing and translations.
Does he use proprietary or open models?
He runs his own GPU and CPU servers with open models, which allows for bulk AI work without total dependence on paid APIs.
What is 'BusinessOS'?
BusinessOS is an AI-native operating system that provides a unified foundation (identity, permissions, and audit logs) for a whole business to replace disconnected tools.
Where is his AI studio located?
The studio operates from Palghar, Maharashtra, and utilizes a team of human operators and a fleet of specialized AI agents.
