Cloudnix · Product studio · since 2011

Product engineering, from architecture to production.

Cloudnix is a product studio. We design, build and operate our own platforms, and take on a small number of AI engineering projects.

Platform · since 2011 Shopnix Multi-tenant ecommerce for Indian businesses
Platform Mynix Configurable workflows with SLAs
Platform Autocon Agentic content pipeline, cross-LLM
Client system · built for Bugasura Bugasura AI RAG and agents for test generation
Client system · built for Moolya Moofu Minesweeper-style testing platform

Businesses that run on what we built

Appachhi
Moolya
Homelane
Sapnaonline
Coorgshoppe
Startuptestlab
Eatrich
Meraki
Happygo
Avanze
2011today
2011Shopnix in production since: 15 years
1000+teams use Bugasura, whose GenAI we built
6platforms built: 4 of our own, 2 for clients
+7
13engineers, designers and testers

Our platforms

Our product · since 2011

Shopnix

Indian businesses sell on their own store and app: UPI and COD, GST invoices, couriers and WhatsApp updates, with 0% commission.

Multi-tenant commerceUPI, cards and CODGST invoicingWhatsApp messaging
Architecture and engineering →
Shopnix
Our product

Mynix

Sales, HR, finance and support run their processes on one platform: boards, forms, workflows with SLAs and dashboards. We run our own company on it.

Configurable workflowsSLA engineForm builderRole-based access
Architecture and engineering →
Mynix
Our product

Autocon

An agentic content pipeline: agents research topics, draft in several tones, review each other across LLM providers, illustrate and publish. A person approves, and each rejection becomes a rule.

Cross-LLM reviewMulti-tone writingHuman-in-the-loopHeadless publishing
Architecture and engineering →
Autocon

Client systems

Built for Bugasura

Bugasura AI

The GenAI layer behind Bugasura's testing agents: a knowledge base of each customer's product, retrieval over it, and swappable LLMs on demand.

RAG over product knowledgeVector searchModel-agnostic LLMsAgentic test generation
Architecture and engineering →
Bugasura AI
Built for Moolya

Moofu

A minesweeper-style testing platform: the app is laid out as a board of areas, testers clear it area by area and flag the bugs they hit, and the board shows what is still unexplored.

Gamified testingCoverage boardBug flaggingDevice and network matrix
Architecture and engineering →
Moofu: the app is laid out as a board of areas; testers clear it area by area and flag bugs where they find them, and coverage shows what is still unexplored. NEW BUILD app v2.4 laid out as a board of app areas THE BOARD Login Search Cart Pay Offers Wallet Profile Orders Track Return Chat Alerts Signup Filters Wishlist Share Address Coupons Reviews Help Rating Settings Refund Logout COVERAGE Areas cleared FLAGS Pay: timeout on 2G Return: wrong total each flag with steps and device

AI engineering, in detail

How we design the AI parts of a system. The same patterns run in our platforms and in the client systems above.

INGEST Sourcesdocs, tickets,database rows Chunk Embed[0.12, −0.48 …] Vector store kept in syncthrough change logs ANSWER Question"refund rules?" Retrievetop matches Rerank LLManswers from context,with citations Evaluategrounded? on-topic?
YOUR ENVIRONMENT · ON-PREMISE OR PRIVATE CLOUD Your appsame API as cloud Gatewayauth, quotas, logs Model server Llama Gemma fine-tuned quantised to fit the hardware GPU Retrievalprivate index Your datadocuments, tickets, recordsnever leave this boundary Cloud LLMsonly if you allow it,for non-sensitive tasks Same code path, soswitching local ↔ cloudis configuration
Requesttask + data Router sensitive data → localsimple → fast and cheaphard → reasoning modelrepeat → cache Local modelprivate data Fast cloud modelhigh volume Reasoning modelhard problems limit or outage →fail over Cost per call
Plannerbreaks the task down Executor searchDBAPI → structured JSON Checker modelanother provider Personapproves or notes Doneto your systems revise REGRESSION EVALS · EVERY RELEASE accuracy grounding format tone safety latency cost
  • Chunking tuned to the source: tickets, docs and database rows are split differently
  • Embeddings in a vector store (Qdrant), kept in sync with the source through change logs
  • Answers grounded in retrieved context, with an evaluation pass for faithfulness
  • Open models (Llama, Gemma and similar) served on your own servers or GPUs
  • Data never leaves your environment: retrieval, inference and logs stay inside
  • Quantised models sized to the hardware, behind the same API as cloud models
  • Each task goes to the model that fits it: local for sensitive data, cloud for heavy reasoning
  • Automatic failover when a provider hits a limit or an outage
  • Caching and a cost per call on every request
  • Agents plan, call tools and return structured output that the system can trust
  • A second model reviews the work; a person approves where it matters
  • Regression evals on every release, so quality does not drift

Products and client work

The same engineers work on both. Operating our own platforms informs the systems we build for clients, and the reverse.

01

Our products

Shopnix, Mynix and Autocon are ours: we build, host, monitor and support them, and carry the cost of every design decision.

02

Client builds

A small number of client systems at a time, built by the same team: the GenAI layer for Bugasura, the testing platform for Moolya.

Engagements

Embedded

AI Native Teams

Engineers who build AI-first, embedded in your team.

Your product AI engineer AI engineer QA engineer Designer
  • Cross-LLM engineering: agents, retrieval and evaluations
  • Production habits: security, cost per call, monitoring and fallbacks
  • A pod of two AI engineers, a QA engineer and a designer
  • Working in your tools and your sprints
End to end

Build, Deploy, Secure, Manage, Transfer

We take an AI system from first design to a clean handover.

  1. BuildArchitecture first, then the system
  2. DeployOn your cloud or ours, monitored from day one
  3. SecureGuardrails, access control, data kept where it belongs
  4. ManageWe operate it, from models and costs to incidents
  5. TransferCode, documentation and runbooks, when you are ready

For funded startups and established companies building AI into their product or operations. Discuss a project →

Engineering principles

01

Foundations first

Anyone can make an app look good. We start with the architecture: the data, the security, how it scales. You can always change a screen. You can't change a foundation.

02

The full team, not just coders

Analysts who understand the business, architects who design it, testers who try to break it, and engineers who keep it running after launch.

03

We run what we build

We patch, monitor and scale our own platforms every day. Shopnix has been live since 2011, and we are still the ones on call.

04

Judge us by our software

Don't go by our website. Open Shopnix or Mynix and use them. That is the quality you get.

Team

Engineers, designers, testers and sales, in front of the wall everyone walks past each morning. The whole team
You can always change your UI. You can never change your app's architecture.
Avinash Nishant
Avinash NishantFounder, Cloudnix
Mukunda Ramesh Mukunda RameshSales Manager
Vishal S Vishal SUX/UI Designer
Avinash KS Avinash KSSenior Tester
Kaviya M Kaviya MMarketing Executive
Jeswin Dass Jeswin DassAssociate UX/UI Designer
Gayathri K Gayathri KAssociate Software Engineer
Jeya Kirubakaran Jeya KirubakaranAssociate Software Engineer
Manoj Kumar Pitla Manoj Kumar PitlaAssociate Software Engineer
Samprit Gouda Samprit GoudaAssociate Software Engineer
Sivakrishna Mukku Sivakrishna MukkuAssociate Software Engineer
Umashankar Neelam Umashankar NeelamAssociate Software Engineer
Upendranath Metikala Upendranath MetikalaAssociate Software Engineer

Discuss a project

Share a short brief. If it is a fit, the next step is a call with one of our engineers.

Colour: Sky cyan