Skip to content
Nimbus Analytics Cloud
Back to Projects
SaaS · Web App

Nimbus Analytics Cloud

A real-time analytics platform processing 4B events a day.

Overview

Nimbus had outgrown its monolithic analytics product. Dashboards lagged by minutes, onboarding new enterprise tenants took weeks, and the engineering team spent more time firefighting than building.

BloomByte partnered with Nimbus to design and deliver a ground-up replatform: a streaming ingestion pipeline, a columnar analytics store, and a modern multi-tenant front end — all without a single minute of customer downtime during cutover.

The Challenge

The existing system batched data every fifteen minutes and buckled under spikes. Enterprise customers — the highest-paying segment — demanded sub-second freshness and strict tenant isolation.

Any new architecture had to migrate years of historical data and run in parallel with the legacy stack until it earned trust.

Our biggest customers were threatening to churn over data latency. We needed a team that could re-architect under pressure without breaking what was already live.
Priya Nadkarni, VP Engineering, Nimbus Data

Our Approach

  1. 01

    Discovery

    Two-week audit of the legacy pipeline, tenant data model, and SLA commitments to map every failure mode.

  2. 02

    Architecture

    Designed a streaming ingestion layer on Kafka feeding a ClickHouse cluster with row-level tenant isolation.

  3. 03

    Build

    Shipped incrementally behind feature flags, validating each subsystem against production traffic shadows.

  4. 04

    Migration

    Dual-wrote to old and new stores, reconciled nightly, and cut tenants over in controlled waves.

  5. 05

    Launch

    Full cutover with automated rollback, real-time observability, and on-call runbooks handed to the Nimbus team.

The Solution

We delivered a horizontally-scalable analytics platform: a Go ingestion service writing to ClickHouse, a Next.js dashboard with server-streamed queries, and a Kubernetes deployment that auto-scales with event volume.

Streaming Ingestion

Kafka-backed pipeline with exactly-once semantics and back-pressure handling.

Sub-second Queries

Columnar storage and materialized views deliver dashboards in under 400ms at p95.

Tenant Isolation

Row-level security and per-tenant resource quotas with zero cross-tenant leakage.

Self-serve Onboarding

New enterprise tenants provisioned in minutes instead of weeks via an automated control plane.

Results & Impact

4B+

Events processed daily

92%

Faster dashboard loads

$3.4M

Expansion revenue unlocked

0

Minutes of cutover downtime

Gallery

BloomByte didn't just rebuild our product — they handed us a platform our team is proud to own. The replatform paid for itself within two quarters.

Priya Nadkarni

VP Engineering, Nimbus Data