Varun Cumbamangalam
ProjectsAboutContact

Varun Cumbamangalam.

Senior IoT and Edge AI Engineer, AI Product Manager at OraLens Healthcare

ProjectsAboutContactPrivacy

© 2026 Varun Cumbamangalam. Built with Next.js and Tailwind CSS.

All Projects
RenderBoard AI - Interior Design Rendering Platform logoRenderBoard AI - Interior Design Rendering Platform logo
AI/MLBackendFrontendDevOps

RenderBoard AI - Interior Design Rendering Platform

From a floor plan to an approved, rendered presentation board, with every paid step gated.

RenderBoard AI - Interior Design Rendering Platform main screen
6

Services

Scale-to-zero

GPU Workers

1,896

Backend Tests

Tech Stack

FastAPICeleryRedisPostgreSQLNext.jsThree.jsBlender CyclesReal-ESRGANSAM 2Azure Container AppsTerraformClamAV

The problem

Interior teams need client-ready options quickly, but AI rendering tools hide provider cost, moderation failures, asset provenance and the point where a human must approve. A rendering pipeline for real client work needs those boundaries in the open.

The challenge

I built one supervised journey: plan upload through a malware-scanning quarantine, floor-plan parsing with a scale and correction step, layout generation, per-room approval, then rendering and board assembly. GPU work (path tracing, upscaling, segmentation) had to scale from zero on Azure without stranding jobs, and every stage had to be provable with gate tests rather than screenshots.

System Design

Web app and API in front of Celery workers, OCR, quarantine, upscale and segmentation services, with GPU render workers that scale from zero.

A monorepo of six services on Azure Container Apps in South India: a Next.js web app, a FastAPI API with Celery workers on Redis, a Tesseract OCR service, a ClamAV upload quarantine, a Real-ESRGAN upscaler and a SAM 2 plus Grounding DINO segmentation service. Postgres and Blob Storage hold state and artifacts. GPU workers run with a minimum of zero replicas and wake from a Redis wake-list. Direct Three.js drives the in-browser preview, with Blender Cycles for final renders, and Terraform describes the infrastructure.

Results

RenderBoard is deployed end to end as an internal test release at renderboard.varunc.com and monitored through Watchdog. A synthetic production journey has taken a plan from upload through room approval to a 4K path-traced render. The backend baseline is 1,896 passing tests, preview-to-Cycles visual similarity scores 0.90 against a 0.85 gate, and multi-view consistency 0.98 against a 0.90 gate.

Gallery & Demos

Plan Review

Plan Review

Walls, rooms and scale confirmed on the uploaded floor plan.

Room Layout

Room Layout

Furniture and finishes placed per room before approval.

Export Review

Export Review

Approved renders assembled for the presentation board.

Product Website

Product Website

The RenderBoard landing page.

Click any image or video to expand · ← → keys navigate

Personal Project

More from Personal Project

Ummidvar - AI Job Application Agent

Autonomous, multi-profile job application agent: discovers roles across 7+ boards, ranks each 0-100 with a 5-component explainable scorer, generates hallucination-safe tailored applications via a Facts Graph, auto-submits through real ATS systems, tracks replies via Gmail, and flags visa-sponsorship signals across 5 countries - 286 tests, 11 modular packages, zero vendor lock-in.

PythonFastAPIPlaywright

Ummidvar - Autonomous Job Agent

Product vision and feature scoping for Ummidvar - an autonomous, multi-profile job application agent. Defined the 11-package modular architecture from a product perspective, prioritised the anti-hallucination Facts Graph as a non-negotiable product constraint, and scoped visa-sponsorship intelligence for 5 international markets.

PRD AuthoringFeature PrioritisationMulti-Persona Design

Sylars Work Manager - Agent Operations Workspace

End-to-end ownership

A self-hosted workspace for assigning repository work to AI agents with exact-SHA project context, isolated task environments, progress steering, and explicit human approval gates.

TypeScriptNode.jsDocker Compose

Watchdog - Multi-cloud Observability Control Plane

End-to-end ownership

A privacy-minimised observability and safe-operations control plane spanning AWS, Azure, GCP, container hosts, registry storage, incidents, and exact human-approved cleanup plans.

TypeScriptNext.jsAWS Lambda

BidWave - Freelance Opportunity Engine

End-to-end ownership

A self-hosted pipeline that scrapes four freelance platforms, de-duplicates listings, ranks fit with a TF-IDF prefilter and an LLM rerank, drafts proposals, and holds every application in a human approval queue.

FastAPISQLAlchemySQLite

LoadCell - Strength Machine Controller

End-to-end ownership

A local-first Android app that drives personally owned Aroleap-compatible strength hardware over Bluetooth LE: set resistance, run custom plans, count reps from the machine's own telemetry, and keep every set even if the process dies.

KotlinJetpack ComposeBluetooth LE

Swadhika Studios - Editorial Interior Portfolio

End-to-end ownership

A calm, editorial website for an interior-design and turnkey-delivery studio, built around full-bleed project photography, deliberate typography, and unhurried motion.

HTMLCSSJavaScript

BidWave - Freelance Opportunity Intelligence

Product Owner

A self-hosted opportunity engine that aggregates freelance listings, ranks fit in two stages, drafts tailored proposals, and keeps application decisions under human control.

Product DiscoveryPRD AuthoringRICE Prioritisation

Atelier CRM - Studio Readiness & Delivery

Product Owner

A status-first CRM for boutique interior-design studios, connecting client approvals, procurement readiness, installation milestones, and project delivery in one operating view.

Workflow ResearchStatus ModelPRD Authoring

Interested in this work?

I can walk through the architecture and code during an interview.

Get in Touch All Projects