I run the full arc — discovery, prototyping, agentic build, delivery, and the team underneath it — across core banking, payments, and capital markets. Most recently: designing and shipping an agentic system that turns raw source schemas into deployable data products.
I've spent my career at the point where banking data breaks — legacy core systems, payment rails mid-migration, risk feeds no one fully trusts. Publicis Sapient has been home since 2016, leading product for global banking clients through core banking migrations, ISO20022 payments modernization, and real-time data platforms.
In the last two years that's shifted from writing roadmaps about AI to building with it. I design and ship agentic systems myself — not just specify them for engineering — because the fastest way to know what an AI-native data platform should do is to build one and watch where it strains.
I'm looking to take on a bigger leadership charter next — owning AI product strategy end to end, prototyping it hands-on, and building the team that scales it.
An agentic workflow that automates the journey from raw source-system schemas to business-ready data products. It generates metadata catalogs, semantic knowledge graphs, and domain models, then produces ODP / FDP designs and deployable PostgreSQL and Neo4j artifacts — collapsing what used to be a manual, multi-week data-modeling exercise into an orchestrated agent pipeline. Designed with dedicated evaluation subflows and guardrails alongside the main pipeline, ensuring output quality and safety at every stage — this is the kind of AI-native data modernization tooling I want to bring into a larger platform team.
Smaller builds where I'm getting hands dirty with agentic systems, LLM tooling, and data patterns — code at github.com/rsri24.
A working model of the end-to-end FX trade lifecycle, from execution through settlement.
View repo → HTMLA demonstration of Data Vault modeling patterns applied to core banking data.
View repo → HTMLAn AI-assisted tool for generating structured questions from source content.
View repo → PythonExploring detection methods for hallucinated output in LLM responses.
View repo → JavaScriptA security-focused analysis tool built to inspect and surface risk signals.
View repo → TypeScriptA conversational assistant for learning Kannada, built and deployed on Replit.
View repo →On building product in regulated, high-stakes banking environments.
Read on LinkedIn → AI · GenAIThoughts on where agentic AI is actually changing how product gets built.
Read on LinkedIn → Markets · InvestingA take on the SpaceX IPO and what it signals for private markets.
Read on LinkedIn →