At Grismo Solutions, I lead the AI team and design prompts, agent behaviour, and delivery for enterprise LLM products used across India, UAE, and Saudi Arabia. My prompts and JSON schemas support production agents that execute financial transactions, where malformed output can post incorrect data.
I rebuilt prompt-context binding for an Oracle AI Agent Studio invoice agent, reducing the workflow from multiple turns to a single user turn. The agent extracts invoice data into a validated schema and posts a PO-matched draft invoice, routing failures to a structured exception.
On the Segregation of Duties Access Governance Agent project, I built an analysis engine and conversational agent over role-privilege records. I grounded responses in a structured payload to prevent fabricated figures and role names, and caught a correctness flaw before release.
Previously, at Riskbricks, I built RAG pipelines for threat-intelligence retrieval and deployed AI applications for cybersecurity workflows. At CodeSentinel, I contributed to the AI and detection design and early product build for a system that monitored application and system logs.
