AK
WorkExperienceCase StudiesTimelineWritingContact
Resume
AK

Case Study

Holocomm — Multichannel AI Booking Agent

AI-powered booking agent that lets customers book through natural conversation across WhatsApp, Telegram, Instagram, and more.

Co-founder building a production-ready multichannel booking platform for fitness studios and service businesses, reaching customers across WhatsApp, Telegram, Instagram, and more via Zernio's unified messaging API.

Founder ProjectFitness / Service BusinessesMVP

Holocomm - Multichannel AI Booking Agent

Building a multichannel booking platform as co-founder. Designed the architecture, led development, and shaped product direction for an AI-powered system that handles natural language bookings across WhatsApp, Telegram, Instagram, and more via Zernio's unified messaging API.

RoleCo-Founder & Lead Developer
Team3-person founding team
Timeline~1 year of active development
IndustryFitness / Service Businesses

At a glance

By the numbers

~70%
turns with zero LLM calls

regex-first bilingual routing

~284K
lines of code

57 modules

8,700
test & eval cases

1,700 fixtures

~21
agent tools

Zod-validated I/O

204 / 214
ports / adapters

hexagonal architecture

What we were solving

Context & problem

Small service businesses like fitness studios and salons juggle bookings across messaging apps, calendars, and payment links. As volume increases, missed messages, double bookings, and manual follow-ups become the norm.

Customers want to book the way they already communicate — on WhatsApp, Telegram, Instagram, or wherever they already chat. But building a conversational booking experience that handles natural language, manages capacity in real-time, and prevents race conditions is a complex engineering challenge.

How we approached it

Solution

We built a multichannel booking experience where customers book using natural language like "I want yoga tomorrow at 7pm" — on WhatsApp, Telegram, Instagram, and more via Zernio's unified messaging API. The AI layer uses LangGraph and LangChain with a ports and adapters architecture - keeping model, tool, and provider concerns isolated as the product evolves. Typed tools handle session discovery, booking management, crew info, locations, memberships, preferences, and business operations.

The NestJS backend uses MongoDB with atomic operations to ensure race-condition safe bookings. Real-time capacity management includes waitlist support with automatic promotion when spots open up. The multi-tenant architecture supports multiple businesses with complete data isolation.

Impact

Outcomes

  • - Broad unit, integration, E2E, and eval coverage across the API, AI flows, and dashboard surfaces.
  • - Per-channel latency budgets, tool timeouts, retries, and circuit breakers keep conversational turns bounded across channels.
  • - Race-condition safe: 50 parallel booking requests on a 10-capacity session → exactly 10 succeed.
  • - Redis-backed caching, MongoDB indexes, and an outbox-style notification path support operational reliability.
  • - Multi-tenant architecture supporting multiple businesses with complete data isolation.
Holocomm scheduling dashboard
SchedulingScreenshot
Booking flowAnimated flow
Holocomm customer profiles
Customer profilesAnimated flow

Behind the scenes

Tech & delivery

Stack

  • NestJS 11
  • MongoDB
  • Redis
  • LangGraph
  • LangChain
  • Zernio Messaging API
  • TypeScript

Challenges

  • Designing a LangGraph tool orchestration layer with ports and adapters around model, provider, and domain boundaries.
  • Implementing race-condition safe booking with atomic MongoDB operations and proper concurrency handling.
  • Building a reliable AI tool execution framework with parallel processing, preflight guards, timeouts, and error recovery.
  • Managing webhook security with signature validation, rate limiting, and idempotency.

How I worked

  • Collaborated with co-founders on product direction and conversational UX patterns across channels.
  • Built comprehensive test suite including E2E chat-to-booking flow tests.
  • Implemented production monitoring with performance metrics and circuit breakers.

What's next

Roadmap

Phase 1

Rich Media Support

Session images, voice transcription, PDF receipts

Phase 2

Smart Recommendations

Personalized session suggestions based on booking history

Phase 3

Business Analytics

Real-time occupancy rates, revenue optimization, demand forecasting

Holocomm: Multichannel AI Booking

Natural language booking across channels — WhatsApp, Telegram, Instagram, and more via Zernio's unified messaging API — powered by AI and built for reliability at scale.

Conversational, on any channel

Book classes with natural language like "I want yoga tomorrow at 7pm" - the AI understands context and handles the entire flow through conversation.

9:41

Holocomm AI

online

I want yoga tomorrow at 7pm

10:30

Found Yoga class tomorrow at 7:00 PM with Sarah. 3 spots left. Book?

10:30

Yes please

10:30
Message

AI Tool Execution

LangGraph and LangChain orchestration with typed tools for session discovery, bookings, crew, locations, memberships, and business details.

tool_registry.ts
>findSessions()✓
>list_sessions()✓
>createBooking()✓
>cancelBooking()✓
>findUserBookings()✓
>findCrew()✓
>getBusinessInfo()✓
>listLocations()✓
>▌

Atomic Booking Operations

MongoDB atomic updates and guarded booking writes help prevent double-bookings under concurrent demand.

race_test.log

// Booked successfully

// Rejected (capacity full)

10/50 booked

Smart Session Management

Dynamic pricing, real-time capacity tracking, waitlist with automatic promotion, and configurable booking deadlines.

session_dashboard.tsx

Yoga

7/10$25

Pilates

FULL$30

HIIT

4/12$20

// Waitlist queue

3 waiting+1 promoted!

Flexible Payment System

Multiple payment methods (card, cash, credits, membership), time-based refund policies (full > 24h, half > 12h, none < 2h), and complete payment lifecycle tracking.

refund_policy.ts
100%
>24h refund
50%
>12h refund
0%
<2h refund
24h+12h2h0h