Let's talk

From Markets to Agents

The AI Startup Founder from India Story

After a decade in the high-stakes Gold and FX markets, Divyaprakash Jha transitioned from building automated trading systems to founding two AI companies in 2026.

By Divyaprakash JhaUpdated 5 min readTopic: Founder

Photo for The AI Startup Founder from India Story

From Trading Systems to AI Agents

Divyaprakash Jha is an Indian AI founder who transitioned from building automated trading systems for high-stakes markets to founding GodEngine and AskShiva. He now operates a specialized AI studio from Palghar, Maharashtra, using a hybrid model of human operators and specialized AI agents.

The journey began in 2016 when Divyaprakash Jha entered the Gold and FX markets. These markets are famously unforgiving; they do not care about confidence, only whether a system holds up when market conditions shift. This decade of experience taught a fundamental lesson in discipline and risk management that now drives his AI development.

During this period, he built over 500 automated trading systems, known as Expert Advisors (EAs), for the MetaTrader 5 platform. These systems were designed to execute specific rules with mathematical discipline, acting as small, autonomous agents that read signals and manage risk without a human needing to click a button.

Founding GodEngine and AskShiva in 2026

In April 2026, Divyaprakash Jha went fully independent to focus on building AI full-time. This move led to the establishment of two distinct private limited companies: GodEngine and AskShiva.

The shift in focus was a natural evolution. While his previous work involved executing rigid trading rules, his new ventures focus on building decision intelligence and task-oriented assistants. He moved from building systems that simply follow instructions to building systems that can reason, remember, and complete complex workflows.

His work is rooted in the local ecosystem, operating from Palghar, but designed for a global standard of technical excellence. He has moved from serving individual traders to building platforms that can power entire organizations.

GodEngine: Decision Intelligence for High-Stakes Choices

Most AI tools today are designed to answer questions, often providing fluent but potentially false information. GodEngine was built to solve a different problem: helping people make difficult decisions.

A real decision involves moving parts: what is true right now, what causes what, what could happen next, and—most importantly—what is the cost of being wrong? Instead of a single chat reply, GodEngine treats a decision as a model of the world. It uses a "resolve pipeline" to break down a question into its core components.

How GodEngine Works

  • Graph Entities: It extracts facts and cause-and-effect links from your question and places them in a clickable knowledge graph.
  • Scenario Probing: It builds plausible scenarios and probes them to see what would have to be true for a specific outcome to occur.
  • Auditable Reasoning: Every claim in a report points back to a specific source or entity, allowing you to audit the reasoning rather than just trusting the text.

This makes it an essential tool for founders choosing between markets or hires, and for analysts who need a defensible, structured write-up rather than a single, unverified number.

AskShiva: Bridging the Gap Between Answer and Done

For many small business owners, a clever AI paragraph is not enough. A shop owner in India doesn't just want to know *how* to make an invoice; they want the invoice created, formatted correctly, and sent to the customer in their local language.

AskShiva is designed to bridge this gap between providing an answer and completing a task. It is built specifically for the Indian business context, where workflows often revolve around WhatsApp, GST, and multiple regional languages.

Feature Standard Chatbots AskShiva
Primary Goal | Providing answers/text | Completing tasks ("Done")
Context | Often loses thread | Maintains memory across conversations |
Local Integration | General purpose | Built for WhatsApp, GST, and UPI |
Language | English-centric | Supports Marathi, Hindi, and more |

AskShiva can handle multi-step flows, such as drafting a GST bill for a specific amount and then preparing a WhatsApp message in Marathi to send to the client. It treats Indian business documents as first-class citizens in its workflow.

A New Way to Build: Three Humans, Fifteen Agents

The company operates under a unique "studio model." Rather than a large traditional agency, the team consists of just three human operators directing a fleet of fifteen specialized AI agents. This allows the studio to ship products much faster than a traditional large-scale development firm.

The agents are not just generic bots; they are given specific personas and roles, often named after Sanskrit or Hindi terms to reflect their craft:

  • Vastu: Architecture and structure.
  • Sutra: Coding and logic.
  • Akshar: Lettering and text.
  • Vani: Words and communication.
  • Ganit: Numbers and mathematics.
  • Parakh: Testing and verification.

The workflow follows a "plan, act, verify" loop. Humans define the high-level intent and the constraints, while the agents work in parallel to build, test, and refine the product. This ensures that every idea is "attacked" early in the process to eliminate weak concepts before they consume significant resources.

Owning the Stack: Local Hardware and Open Models

To maintain independence and control costs, the studio does not rely solely on expensive, third-party APIs. Instead, they run their own GPU and CPU servers using open-source models. This approach is vital for handling bulk AI work without the unpredictable costs of paid providers.

The technical architecture is built for production reliability. This includes:

  1. Token-based execution: Ensuring that AI runs are bounded by budgets to prevent infinite loops or runaway costs.
  2. Checkpointing: The system saves progress after every step, allowing a process to resume even if a server restarts.
  3. Real-time streaming: Using server-sent events so users can watch the AI's reasoning happen in real-time, rather than waiting for a single large block of text to appear.

BusinessOS: The AI-Native Operating System

Most Indian SMEs suffer from fragmented data. They might use one app for WhatsApp, another for GST, a spreadsheet for HR, and email for invoices. Because these tools don't talk to each other, it is nearly impossible to get a clear view of the business.

BusinessOS aims to solve this by providing a single, AI-native operating system. It is built on a foundation of identity, permissions, and audit logs. Rather than bolting features onto a platform, it builds a secure, multi-tenant environment where every module shares the same data and security rules.

The goal is to move away from "disconnected tools" and toward a unified workspace where AI can actually be useful because it has access to the entire business context—from customer leads to payroll and inventory.

Frequently Asked Questions

What kind of AI companies did Divyaprakash Jha found?
He founded GodEngine, a decision intelligence engine, and AskShiva, an AI assistant designed for Indian business workflows.
How does the AI studio operate?
The studio uses a hybrid model of three human operators and fifteen specialized AI agents to build and ship software.
What makes AskShiva different from a standard chatbot?
AskShiva is built for "done" rather than just "answers," capable of creating GST invoices and managing tasks in Indian languages.
What is the core philosophy behind GodEngine?
It focuses on decision intelligence by providing structured reasoning, causal links, and scenario probing rather than just fluent text.
Does the founder use paid APIs for all AI work?
No, the studio runs its own GPU and CPU servers with open models to handle bulk AI work independently.

Questions people ask

What kind of AI companies did Divyaprakash Jha found?

He founded GodEngine, a decision intelligence engine, and AskShiva, an AI assistant designed for Indian business workflows.

How does the AI studio operate?

The studio uses a hybrid model of three human operators and fifteen specialized AI agents to build and ship software.

What makes AskShiva different from a standard chatbot?

AskShiva is built for 'done' rather than just 'answers,' capable of creating GST invoices and managing tasks in Indian languages.

What is the core philosophy behind GodEngine?

It focuses on decision intelligence by providing structured reasoning, causal links, and scenario probing rather than just fluent text.

Does the founder use paid APIs for all AI work?

No, the studio runs its own GPU and CPU servers with open models to handle bulk AI work independently.