I build the automation that runs my own business.
I'm a registered dietitian who got tired of doing operations by hand, so I learned to build the systems instead. Health 7 — my 1:1 coaching practice — runs on a self-hosted n8n stack I designed: lead capture, CRM, AI voice qualification, WhatsApp, and approval gates before anything reaches a human. Everything below is live, not a course project.
Don't take my word for it
The canvas and the code, so you can judge the work before we ever speak.
Watch the funnel run
A short walkthrough of the live n8n canvas — a lead comes in, the router fires, the approval lands in Slack, and the message goes out. No slides.
Read the workflows
Sanitised exports of the pipelines above — node structure, branching logic, and payload shapes, with credentials and customer data stripped out.
Systems, not experiments
Each of these is running in production for Health 7 or a client. Happy to walk through any of them live.
Lead funnel & CRM automation
Website enquiry hits a webhook, lands in a structured CRM, and triggers geo-aware follow-up — different sequences for India and overseas leads. Slack approval gate sits between the draft and the send, so nothing goes out unreviewed.
AI voice qualification campaign
An outbound voice agent calls inbound leads, runs a structured health-goal conversation, and returns typed variables — intent level, goal, callback window. A router branches on those into hot, warm, and retry paths. Built the WhatsApp Business API integration end to end, including Meta verification.
Multi-agent orchestration layer
Eight specialised Claude Code agents covering operations, content, and clinical plan drafting, wired to a version-controlled knowledge base. It produces a morning brief in Slack every weekday without me asking.
Application tailoring engine
Paste a job posting, and a three-pass LLM chain rewrites my CV against it, renders a formatted PDF, posts it to Slack for approval, then sends and logs the application. Deliberately no auto-submit — the human stays in the loop on anything that represents me.
Real-time voice assistants
Two desktop assistants that take spoken instructions and act. The interesting problem was latency: I traced a ten-second response delay to a model config setting and brought it down to about one second. Also killed a console-spawn bug across ~160 subprocess calls.
Website and storefront
Designed and shipped health7.co.in, plus a working supplement storefront with a Postgres backend and order webhooks that fire into the same automation stack. Checkout and post-purchase flows built from scratch, not assembled from a theme.
Scripted video production
Turned reel editing into a pipeline: automatic multi-take trimming, speech-to-text captioning, branded overlays, and music, all driven by scripts rather than a timeline. Cut edit time from hours to minutes.
Keeping it all running
I own the servers too. Docker deployments, DNS, SSH hardening and key rotation — and a week-long outage I diagnosed down to the virtualisation layer while keeping the business moving. I've debugged this stack at 3am, which is a different skill from building it.
Stack
Listed by what I've actually shipped with, not what I've read about.
Background
I trained as a registered dietitian and spent my early career in hospital dietetics and then at Healthify, running behaviour-change coaching for premium members and watching where the operations broke.
The pattern was always the same: good clinical work drowning in manual follow-up. So I started building the missing layer myself — first for my own practice, then as the thing I'm best at.
That combination is the point. I can read a health product's user journey as someone who has sat across from the patient, and then go build the routing, the integrations, and the infrastructure that makes it work at volume. Most people bring one half of that.
Currently founder of Health 7 and founding behavioural science lead at an AI heart-health platform. Completing an M.Sc. in Dietetics & Food Service Management.