AI Transformation Projects

Contributions as a Google Developer Expert

1. Planet AI — Climate Warning System in Vernacular Languages

Planet AI grew out of a simple observation: the communities most exposed to floods, cyclones, and heatwaves in India are often the last to receive a warning they can actually understand. I worked on building an AI system that takes climate and weather signal data and turns it into clear, actionable alerts delivered in local vernacular languages rather than English or Hindi-only bulletins. The bigger design challenge wasn’t the forecasting itself but the translation layer — making sure a warning in Odia, Tamil, or Marathi carried the same urgency and precision as the source data. For rural and semi-urban India, where smartphone penetration has outpaced formal disaster literacy, this kind of last-mile localization is often the difference between a warning being read and a warning being acted on. It’s a reminder that AI transformation in climate resilience isn’t just a modeling problem, it’s a language and trust problem. Projects like this show how GCC-style engineering talent in India can build solutions that serve India first, rather than importing global platforms that assume English fluency.

2. Nariswatha — Vernacular Women’s Health Platform

Nariswatha was built around a gap I kept seeing in women’s healthcare across India: proactive care is rare when the information women need isn’t available in their own language or in a format they’re comfortable engaging with. The platform uses AI to surface preventive health guidance, symptom triage, and care pathways in vernacular languages, aiming to shift the pattern from reactive treatment to earlier intervention. A lot of the effort went into making the experience feel private and judgment-free, since health-seeking behavior for women in many parts of India is still shaped by stigma and lack of accessible information. Getting the tone and cultural context right mattered as much as the underlying AI accuracy. At scale, this kind of platform has real potential to move the needle on maternal and women’s health outcomes in Tier 2 and Tier 3 India, where specialist access is limited. It’s the kind of project that reinforces why I care about applying agentic AI to problems with genuine social return, not just enterprise efficiency.

3. Prescription & Discharge Summary Simplification in Vernacular Languages

Anyone who has walked out of an Indian hospital with a discharge summary full of clinical shorthand knows how little of it a patient or family actually understands. I contributed to a summarization system that takes prescriptions and discharge notes and rewrites them in plain, vernacular language, keeping the clinical accuracy intact while stripping away the jargon that creates confusion after a hospital visit. The hardest part of this work was calibrating the model so it never oversimplified to the point of losing critical instructions, like dosage timing or warning signs to watch for. Getting that balance right required close collaboration with clinicians, not just prompt engineering. For a country where health literacy varies enormously across regions and languages, this kind of tool reduces medication errors and missed follow-ups in a very direct, measurable way. It’s a small piece of the puzzle, but it’s the kind of AI transformation that touches ordinary patients rather than just hospital back offices.

4. USG Recording Automation with Human-in-the-Loop Feedback

Ultrasound reporting in most Indian hospitals still leans heavily on manual dictation and transcription, which eats into radiologist time and introduces avoidable delays. The project I worked on used voice-driven capture during USG scans, paired with a human-in-the-loop feedback layer so radiologists could correct and refine the AI’s output rather than trust it blindly. That human oversight step was non-negotiable given the clinical stakes involved — the goal was never to replace radiologist judgment but to remove the repetitive transcription burden around it. Rolling this out meant working closely with hospital operations teams to fit the tool into existing scan-room workflows without disrupting patient throughput. The efficiency gains showed up quickly in reduced reporting turnaround time and fewer transcription errors reaching patient files. For large hospital networks operating at scale across India, that kind of workflow efficiency compounds into meaningfully better capacity utilization without adding headcount.

5. Automated Agentic Testing Framework with SDLC Governance

This project sits closer to my enterprise AI engineering work — building an agentic testing framework that governs how AI-driven test generation plugs into the software development lifecycle, rather than treating it as a bolt-on tool. The framework enforces structure and traceability across test generation, so enterprises get the speed benefits of agentic AI without losing the governance and audit trail that regulated industries require. A lot of the design thinking went into making the framework repo-agnostic, so it could sit across different codebases and testing stacks without needing to be rebuilt each time. For large enterprises building or scaling GCCs in India, this kind of governed automation is what makes agentic AI adoption defensible to risk and compliance teams, not just attractive to engineering teams. It’s one of the clearer examples I’ve worked on of AI transformation that has to satisfy both innovation and control simultaneously. That tension, more than the AI itself, is usually what determines whether these frameworks actually get adopted at enterprise scale.

6. Supply Chain AI Transformation — Agentic Procurement and Ordering

The supply chain work centered on applying agentic AI to procurement and ordering workflows that were previously bogged down by manual review, duplicate purchase orders, and slow vendor coordination. Agents were designed to flag duplicate items before they became duplicate spend, route orders intelligently, and compress the time between a procurement request and actual delivery. The interesting engineering challenge wasn’t the agent logic itself but getting it to work reliably against messy, inconsistent legacy procurement data, which is the norm rather than the exception in most large enterprises. Once that data layer was stabilized, the speed and accuracy gains in order processing became very visible to procurement teams almost immediately. For large enterprises, this kind of transformation translates directly into working capital efficiency and fewer costly duplicate purchases, which finance leadership tends to notice quickly. It’s a good example of how agentic AI delivers value in supply chain not through a single dramatic automation, but through a series of small, compounding corrections across the ordering pipeline.

Published Books

Agentic E-Commerce

This first volume lays out how agentic AI reshapes the e-commerce stack end to end — from discovery and personalization through checkout and post-purchase servicing. I wrote it to give practitioners a concrete architectural vocabulary for multi-agent commerce systems, rather than another high-level trend piece. Agenticera Press, 2026.

Engineering, Industrialisation & Scale

The second volume is the deep engineering companion — how you take an agentic e-commerce system from a working prototype to something that survives production traffic, on-call rotations, and real SLAs. It covers orchestration patterns, observability, and the platform discipline that most agentic AI writing skips over. Agenticera Press, 2026.

Responsible & Sustainable AI-Driven E-Commerce

This volume tackles the parts of agentic commerce that don’t show up in a demo — fairness in recommendation and pricing, environmental cost of always-on agent orchestration, and how to build sustainability into the architecture rather than bolt it on afterward. Agenticera Press, 2026.

Responsible AI Across Emerging & Regulated Industries

The fourth volume takes the responsible AI framework beyond e-commerce into banking, healthcare, and other regulated sectors, where governance isn’t optional. It draws directly on my work building AI governance and risk frameworks inside regulated enterprises. Agenticera Press, 2026.

Enterprise Knowledge Intelligence: Strategy, Architecture & AI-Powered Discovery for Humans and Machines

This book is about the layer underneath every agentic AI system that nobody wants to talk about — enterprise knowledge architecture. It covers how to design knowledge discovery systems that serve both human users and autonomous agents without one degrading the other.

Platform and Model Design for Responsible AI (Packt) &

Revolutionizing Youth Mental Health with Ethical AI (Apress)

Two earlier books round out the set — one on platform and model design practices for responsible AI, published by Packt, and one on applying ethical AI specifically to youth mental health, published by Apress. Both came out of work I’d done in regulated and sensitive-domain AI before agentic systems became the industry’s main focus.