Knowledge Concierge

The AI-Powered Tenant Onboarding Orchestrator

Mayank Tamakuwala • Intern Project • Summer 2026

↓ scroll to begin

The problem

The Problem

Teams want their documents searchable by AI. But setting that up means manually configuring 3 different services, each with its own API, its own format, its own team.

"I just want my docs processed... why is this so hard?"
Service complexity

3 Services, 3 Headaches

KI : fetches your documents on a schedule

KV : processes them through pipelines

KR : makes them searchable by AI

Hours of setup. Dozens of fields. Easy to get wrong.

loading animation...
from chaos → to clarity
KC enters
One Interface.

Meet Knowledge Concierge

Instead of 3 services, users talk to one AI agent. It asks the right questions, fills in the blanks, and sets everything up automatically.

No more config files. No more back-and-forth. Just a conversation.

Under the Hood

KC orchestrates KI, KV, and KR behind a single API. Users never see the complexity.

Architecture
Agent chat

A Guided Conversation

The AI agent walks users through onboarding step by step:

1Who are you? (team & project)
2Where are your docs? (data source)
3How should we process them? (pipeline)
4Done! Project created.
loading animation...
ask → configure → done
What I built

What I Built

•CreateProject & UpdateProject APIs
•Pipeline config lifecycle (draft → publish → active)
•AI agent onboarding flow & guardrails
•Project activation sequence
•CloudWatch observability (dashboard + alarms)
Observability

We Can See It Working

Real-time monitoring of every API call:

• Faults by operation

• Latency (P50 / P99)

• Invocation counts

• Alarms that page the team if something breaks

loading animation...
system goes live

Impact

3 → 1
Services to interact with
~5 min
Onboarding time (was hours)
0
Manual config steps for users
6
APIs shipped
Victory

Thank You!

Mayank Tamakuwala • DOTS-NEST • Summer 2026

Questions?