Skip to content

AlphaVenue.ai

Multi-tenant SaaS for wedding and event venues, with ELLA, its LLM assistant.

Sole engineer · Kindwell Solutions · 2026 – now

104
operational tools
132
data models
179
pages
1,450
test files

How it is built, step by step

  1. Step 1

    The platform

    A React app on an Express/TypeScript API over MongoDB, with Redis and BullMQ workers for everything that runs in the background.

  2. Step 2

    ELLA, with a person in the loop

    The assistant calls 104 operational tools. Reads run; writes stop at a human who approves them.

  3. Step 3

    Grounded answers

    Retrieval over each venue's own knowledge base in Qdrant, so ELLA answers from that venue's documents.

  4. Step 4

    Open to other agents

    An MCP server exposes role-scoped tools to external agents over OAuth.

  5. Step 5

    Voice, live

    Deepgram to an LLM to Cartesia over Twilio, monitored as it happens.

  6. Step 6

    Kept honest

    About 1,450 test files in sharded CI on every change.

AlphaVenue.ai: system sketchtool callsapproves writesretrievalReact web app179 pagesExpress API132 data modelsMongoDBWorkersRedis / BullMQELLA104 toolsQdrantper-venue RAGMCP serverexternal agentsVoice agentDeepgram → LLM → CartesiaTwilioGitHub Actions CI~1,450 test filesevery write waits for a person

In detail

I built the platform end to end and I run it alone: the API, the web app, the workers, the tests, the CI, and the AI layer on top.

AlphaVenue.ai is a multi-tenant platform for wedding and event venues. The Express/TypeScript API sits on 132 MongoDB data models; the React web app has 179 pages; Redis and BullMQ run the background workers; and about 1,450 automated test files run in sharded GitHub Actions CI on every change.

ELLA is the platform's LLM assistant. It calls 104 operational tools through OpenAI tool calling, and every write action waits for a human to approve it. Retrieval-augmented generation runs over each venue's own knowledge base in Qdrant, so answers are grounded in that venue's documents, never the whole tenant pool.

An MCP server exposes role-scoped tools to external AI agents over OAuth, so a customer's own agents can act on the platform within the permissions of the user who authorised them.

The real-time voice agent runs on Pipecat: Deepgram speech-to-text, an LLM, then Cartesia text-to-speech over Twilio, with live call monitoring. Alongside it I replaced GoHighLevel with a native omnichannel CRM for voice, SMS and email.

Next: LogicOne Dialer

Write to me about a role

umer.jahangier@gmail.com

Subject: Role: let's talk

A line about the team and the role is enough; I reply within a day.

Open your mail app