Skip to content
MB

Multi-Agent AI Platform

Studio Operations, Automated

A Slack-native, serverless platform where four cooperating agents run studio operations end to end.

Role

Senior Frontend Developer — full-stack owner

Company

Crescentic

Period

Jan 2026 — Present

01

Fig. — generative stand-in

2026

Built with

TypeScriptVercel AI SDKAnthropic Claudepgvector / PostgreSQLDrizzle ORMCloudflare Workers

Context

Operations that lived in Slack threads

Studio operations ran on human relay — a request in one channel, an approval in another, a spreadsheet somewhere in between. The work was legible to the people doing it and invisible to everyone else.
The brief was to automate it where it already happened, without asking anyone to learn a new tool.

Approach

Four agents, one tool-calling loop

I built and shipped the platform end to end on the Vercel AI SDK, engineering the tool-calling loop and the supporting interfaces across four cooperating agents.
Event-driven automation ties Slack, GitHub webhooks and scheduled jobs together on Cloudflare Workers, backed by a sixteen-table PostgreSQL schema modelled in Drizzle ORM with pgvector for retrieval.
Every surface an operator touches — request, approval, handoff, escalation — is Slack-native. No second destination to check.

Outcome

Observable by default

Usage and cost-reporting dashboards were instrumented alongside the agents rather than bolted on, so platform behaviour and platform spend are inspectable from the first run.

00

Cooperating agents

00

Table Postgres schema

E2E

Shipped solo, front to back

Agent console — tool-call timeline and run inspector
Slack-native surfaces: request, approval, handoff
Usage and cost reporting dashboard

Open to senior frontend & full-stack roles

murtaza.bohra8999@gmail.com