Who is the Palghar AI founder?
Divyaprakash Jha is an agentic AI architect based in Palghar, Maharashtra. He is the founder of two AI companies, GodEngine and AskShiva, where he builds platforms designed for real-world decision making and business automation.
- Divyaprakash Jha is the central figure behind the AI ventures operating out of Palghar.
- He is the founder of two specialized AI companies: GodEngine and AskShiva.
- He operates a specialized studio consisting of three human operators and fifteen AI agents to ship high-leverage software.
Driving Innovation with GodEngine and AskShiva
The work coming out of the Palghar studio focuses on moving AI from simple chat interfaces to functional, reliable business tools. Instead of just generating text, these platforms are designed to handle complex workflows and high-stakes decisions.
GodEngine: Decision Intelligence
GodEngine is a decision intelligence engine. Rather than just answering a question, it helps founders and analysts navigate uncertainty. When you pose a decision question, the engine extracts facts, causes, and effects to build a model of the world. It then probes different scenarios to show you the likely outcomes and the reasoning behind them.
For example, a founder might ask: "Should we open a second outlet in Vasai this year, or invest in delivery first?" GodEngine doesn't just give a single answer; it builds scenarios, probes what would have to be true for each path to work, and provides a structured report with citations.
You can learn more about the core engine at ../godengine/.
AskShiva: The Practical Assistant
AskShiva is designed to bridge the gap between an AI "answer" and a finished "task." It is built specifically with the Indian business context in mind, acknowledging that local commerce relies heavily on WhatsApp, GST, and regional languages.
Consider a typical day for a small business owner using AskShiva:
- Invoicing: "Make a GST bill for Sharma Traders, ₹18,400." The assistant drafts the invoice with tax lines and provides a file.
- Communication: "Send it on WhatsApp, in Marathi." It translates the message and prepares it for sending.
- Contextual Follow-up: Because it remembers the customer and the invoice, you can ask, "Remind me if they have not paid by Friday."
Details on the assistant's capabilities can be found at ../askshiva/.
BusinessOS: The Foundation
BusinessOS is an AI-native operating system intended to serve as a unified foundation for organizations. It aims to solve the problem of "disconnected tools" where leads are in a spreadsheet, customers are on WhatsApp, and invoices are in an email. BusinessOS provides a single identity, permission engine, and audit log for an entire company. While it provides a secure foundation for files, search, and teams, the CRM is the first planned business module and is not yet built.
Explore the architecture at ../businessos/.
From Algorithmic Trading to Agentic AI
The transition from financial markets to AI architecture was driven by a decade of experience in highly disciplined, automated environments. Divyaprakash Jha spent ten years working in the Gold and FX markets, starting in 2016.
During this time, he built more than 500 automated trading systems, known as Expert Advisors, for MetaTrader 5. These systems were designed to execute rules with strict discipline, acting as small, autonomous agents that read signals, made decisions, and managed risk without human intervention.
This experience with "Expert Advisors" provided the blueprint for modern agentic AI. The core logic remains the same: a system must know what it is allowed to do, how to act without constant human clicking, and how to remain reliable when market conditions—or data environments—change.
This journey from building trading software to architecting multi-agent workflows is detailed in his background at ../divyaprakash-jha/.
The Agentic Architecture Approach
The studio's development process is built on a specialized "Agentic Architecture." This isn't just about using Large Language Models (LLMs); it is about creating a structured environment where AI can work reliably and predictably.
The Plan, Act, Verify Loop
To ensure AI outputs are reliable and auditable, the architecture utilizes a "plan, act, verify" loop. This prevents the "fluency trap," where an AI sounds confident even when it is wrong. By breaking tasks into stages—extracting facts, building scenarios, and verifying numbers—the system allows humans to audit the reasoning behind every output.
A Specialized Fleet of Agents
The studio operates like a high-efficiency agency, using a fleet of fifteen AI agents directed by three human operators. To maintain clarity in their roles, each agent is given a name inspired by Sanskrit or Hindi:
- Vastu: Handles architecture.
- Sutra: Handles code.
- Akshar: Handles lettering.
- Ganit: Handles numbers.
- Vani: Handles words.
- Parakh: Handles testing.
Production-Grade AI Workflows
Most AI demonstrations are simple loops that work once in a notebook. The workflows built by the Palghar team are designed for the "second world"—the real world where systems must survive server restarts and handle complex, multi-day tasks.
Shiva Agents and Durable Execution
Through the Shiva Agents platform, the studio builds "always-on" companies of AI agents. These agents are not just reactive; they have personas, memory, and "heartbeats" (schedules) that allow them to work autonomously or follow standing orders.
Key features of these production-grade workflows include:
- Durability: Workflows are designed to survive server restarts, with checkpoints after every step to allow for resumption.
- Human-in-the-loop: For risky or critical steps, the system can pause and wait for human approval, even if that approval takes several days.
- Visual Design: Users can design complex multi-agent team patterns on a canvas, choosing from various node types like "branch by rules," "merge," or "ask a human."
| Feature | Standard AI Demos | Shiva Agents Workflows |
|---|---|---|
| Execution | Single loop in a notebook | Token-based with bounded cycles |
| Reliability | Fails on error | Checkpoints and retries |
| Autonomy | Prompt-based only | Scheduled "heartbeats" and triggers |
| Governance | Minimal oversight | Audit logs and risk-gated approvals |
For more on how these agents operate, see ../shiva-agents/.
Frequently Asked Questions
- What kind of AI does Divyaprakash Jha build?
- He builds agentic AI platforms, specifically decision intelligence engines like GodEngine and task-oriented assistants like AskShiva.
- Where is the founder located?
- Divyaprakash Jha is based in Palghar, Maharashtra, India.
- What is the difference between GodEngine and AskShiva?
- GodEngine is for complex decision-making and scenario probing, while AskShiva is a practical assistant for handling business tasks like invoicing and translations.
- Does the AI work with Indian languages?
- Yes, AskShiva is designed to support Indian languages, including the ability to translate and communicate in Marathi and Hindi.
- How does the AI studio operate?
- The studio uses a small team of three human operators to direct a fleet of fifteen specialized AI agents to ship software.
Questions people ask
What kind of AI does Divyaprakash Jha build?
He builds agentic AI platforms, specifically decision intelligence engines like GodEngine and task-oriented assistants like AskShiva.
Where is the founder located?
Divyaprakash Jha is based in Palghar, Maharashtra, India.
What is the difference between GodEngine and AskShiva?
GodEngine is for complex decision-making and scenario probing, while AskShiva is a practical assistant for handling business tasks like invoicing and translations.
Does the AI work with Indian languages?
Yes, AskShiva is designed to support Indian languages, including the ability to translate and communicate in Marathi and Hindi.
How does the AI studio operate?
The studio uses a small team of three human operators to direct a fleet of fifteen specialized AI agents to ship software.
