Agentic AI assistants, workflow engines and sync systems that keep messy real-world data in order. Lately that means a Claude-powered assistant on the backend and its Flutter chat on the front.
I like problems where the hard part is keeping everything in sync: many systems, many channels, and a user who should never notice any of it.
I studied Electrical Engineering at IIT Delhi and ended up writing distributed backends. At Accolade I work on a property-management platform: scheduling technicians, running conversations over chat, SMS, email and voice, and syncing residents and leases with property management systems like Rent Manager.
Before that I built compliance automation in Go at Letsbloom, a Standard Chartered Ventures company, and supported Standard Chartered's global rollout of a communications surveillance platform.
The jobs, and the things inside them I'm proudest of. Open any card for the longer story.
Software Engineer
I build the backend of a property-management platform: an agentic AI assistant, Temporal workflows for conversations and data sync, technician scheduling, and the event-driven plumbing that keeps it all consistent with external systems.
An AI assistant built on Spring AI and Claude Managed Agents that finds the right tool for over 30 use cases. I also built the whole Flutter chat: live streaming over AG-UI, and tables, forms and questions the agent can show as cards.
Technician scheduling driven by the Timefold solver. It weighs shifts, travel, priority and permission to enter, then hands the plan to a supervisor to review and approve before anyone sees it.
I redesigned how the platform talks to people. Leads, residents and vendors now share one participant-agnostic conversation model, a Temporal workflow drives every reply, and messages live in an S3-backed store with a local cache instead of the main database.
Two-way sync with Rent Manager and other property systems through Kafka and the transactional outbox pattern, so no change gets lost. Integration tests run against simulated S3 and SQS.
Work orders and lead lifecycles move through strict, matrix-driven transitions. The right side effects follow each move, like assigning a vendor and opening a conversation with them.
Software Developer
Compliance automation at a Standard Chartered Ventures company. I built Go and Temporal services that ingest GitHub organisation data across tenants and run policies on it, plus the portal's first module, authentication.
Development Specialist
Supported the global rollout of the Smarsh communications surveillance platform, working with compliance teams to pin down what each country's rules require.
Zip packaging for audit evidence up to 8 GB. Documents stream from Google Cloud Storage into chunks and back, so memory stays flat, and users download through short-lived signed links.
A deployment system for many AWS accounts at once. A custom state machine keeps each infrastructure rollout atomic, and a worker pool runs them in parallel.
A drag-and-drop holding shelf that floats above every app, so files have somewhere to wait while you switch windows. It has its own Next.js site, too.
Event-driven state management for Go apps. It uses reflection and concurrent observers to keep state transitions type-safe.
A vision pipeline fine-tuned on more than 10,000 images, sped up with ONNX Runtime, plus a PyQt5 tool that pre-labels new images.
B.Tech, Electrical Engineering
Indian Institute of Technology Delhi, where I traded circuits for distributed systems.
A blog is on the way. First up, a few posts I'm drafting:
Open to interesting backend, workflow and agentic AI problems. The fastest way to reach me is email.