I started in machine learning: not because it was fashionable, but because I was obsessed with what you could infer from data that had no obvious structure. At SRM Institute in Chennai, I built a cyclone intensity prediction system using infrared INSAT-3D satellite imagery. The hypothesis: convolutional nets applied to raw geoTIFF frames could yield faster, more granular predictions. It worked, and reinforced a way of thinking I still carry: physical systems as data problems.

Around the same time: real-time object detection on the COCO dataset, BERTopic topic modelling on news corpora, YOLO-based social distance monitoring. Research rigour. Data-to-decision pipelines. Systems thinking.

Then products. An iOS trip planning app (Circles), built in SwiftUI with MVVM architecture and a MongoDB backend. Ten-plus screens, collaborative editing, real-time sync. An iOS internship at Infosys. A hospital management system (zen-u). An open-source 3D slicer fork (slic3r).

The pattern: full-stack capability, product ownership, cross-platform range. The discipline of shipping something complete. The distance between “it works” and “it feels right”, and learning to close it deliberately.

Electric Miles. I joined as a Platform Onboarding Engineer and spent the first phase integrating EV chargers into a live OCPP network: 30+ chargers across 15+ manufacturers. KEBA, Rolec, NexBlue, EnSmart, EN+, Vestel, Heliox, and others. Every manufacturer treats the OCPP standard differently. Custom status codes, non-standard field values, undocumented edge cases. You learn fast that a protocol spec and a real charger are two different things.

Then: full end-to-end test ownership of emPACT, Electric Miles’ flagship B2B platform for CPOs and fleet operators. State machine coverage for every charger type. Payment flow edge cases. Load balancing scenarios.

Now: AI & Backend Engineer. I led the internal rollout of Claude (Anthropic’s API) across Electric Miles’ operations, automating high-friction workflows in the PHP/Symfony backend. The first production AI integration at an EV infrastructure company. That’s the work I find most interesting: the intersection where energy infrastructure meets AI systems.