Serasi

Inspiration

This project was born from our teammate Melissa’s real-life frustration in Indonesia. She witnessed firsthand how many small to medium-sized hair salons still rely heavily on manual booking systems or basic WhatsApp messaging. Customers often have to travel to the salon during peak hours, only to find out there are no available slots. While some salons use WhatsApp for bookings, high traffic volume and overwhelming admin tasks frequently result in slow or no responses. This leads to poor customer experience, wasted trips, double-bookings, and lost business — many customers end up switching to other salons. Although sectors like healthcare (dentists) and some larger beauty chains have adopted more advanced digital tools, the majority of independent hair salons — especially in Indonesia and across Southeast Asia — are still lagging in implementing seamless, intelligent, and personalized booking experiences. Seeing how e-commerce and F&B have been transformed by smooth digital journeys, we wanted to bring that same level of convenience and intelligence to everyday personal services like hair salons.

What it does

Serasi is an intelligent, agentic AI-powered web platform that streamlines the entire hair salon experience for both customers and salon owners. For Customers:

Upload a selfie and our Computer Vision system analyzes your face shape, skin tone, and hair texture.
Our AI Agent instantly recommends personalized hairstyles with virtual try-on previews.
Through a natural conversational interface, the AI guides you from style selection to choosing add-on treatments.
It then matches you with the best available stylist based on their specialties and real-time availability. You can book your appointment seamlessly, leading to fewer unresponsiveness and wasted trips.

For Salons:

Reduces front-desk workload by automating consultations and bookings.
Minimizes scheduling conflicts and no-shows in a trackable system
Helps stylists focus on delivering great services instead of constantly answering messages.

By combining Computer Vision with Agentic AI, Serasi creates a smooth, personalized, and efficient journey; turning a traditionally chaotic process into a delightful digital experience.

How we built it

GroomGrid was built using a modern technology stack. The frontend leverages TanStack to provide a responsive and interactive user experience, while MongoDB serves as the primary database. The backend is implemented in Golang for performance and scalability, and Python is used to develop the AI-powered chatbot and recommendation system.

Challenges we ran into

One of the biggest challenges we faced was integrating and deploying the AI-powered features of our application. During development, the Gemini model frequently returned "busy at capacity" responses, which disrupted testing and development workflows. To maintain reliability, we adapted our implementation by switching from Gemini Flash 3.x to Flash 2.x.

We also encountered changes in Google's AI platform ecosystem. The transition from a postpaid to a prepaid billing model in Google AI Studio required us to rethink our architecture and migrate from the Gemini API to Google Cloud Vertex AI. While this required additional setup and configuration, it ultimately resulted in a more scalable and enterprise-ready solution.

Another challenge involved image generation, where we quickly exhausted the available quota during development and testing. To overcome this limitation, we evaluated alternative models and adjusted our implementation to ensure uninterrupted development.

Lastly, another big challenge our team faced was completing the project while balancing our full-time jobs. Managing our time effectively and coordinating development efforts required significant commitment and discipline. Despite the complexity of the application, we thoroughly enjoyed the process from start to finish. Seeing our vision come to life in the final product gave us a strong sense of accomplishment and pride.

Accomplishments that we're proud of

Six specialized AI agents working harmoniously in accomplishing the user's task Built a full multi-agent pipeline from face analysis all the way to booking confirmation: face_analysis_agent → haircut_options_agent → hairstyle_visualizer_agent → hairstylist_recommender_agent → booking_agent; each with distinct responsibilities, instruction files, and language-aware behaviour.
Meaningful MCP integration at the core of the booking flow MongoDB MCP powers real-time slot availability queries and booking record creation directly inside booking_agent. The agent actually uses the tool to read and write data, making the booking a real persistent transaction rather than a simulated one.
Hairstyle visualization with AI image grounding Built a working visualizer agent that generates reference image queries structured to minimize hallucination and mismatch with a Gemini Vision validation layer as a quality gate before showing results to customers.
Addressing an underserved market in Southeast Asia Independent salons in Indonesia largely rely on informal tools — WhatsApp messages, phone calls, and manual notebooks — for scheduling. According to BPS, over 90% of Indonesian UMKMs still operate without digital systems, and fewer than 10–12% use technology for even basic financial recordkeeping. Serasi AI directly targets this gap: by automating booking, AI-powered consultation, and administrative workflows end-to-end, it reduces manual errors and operational overhead — consistent with findings by Prihandono & Wijaya (2024) that digital transformation significantly improves operational efficiency and data accuracy for Indonesian UMKMs previously dependent on manual processes. Serasi AI is designed from the ground up for this segment, with Bahasa-first interaction and SME-scale operations in mind.

What we learned

Throughout this project, we gained valuable experience in building and deploying a full-stack application that combines AI, mapping services, and real-time user interactions. We learned how to integrate and manage cloud-based AI services through Google Cloud Vertex AI, handle deployment challenges, and coordinate development effectively while balancing our full-time jobs. Most importantly, we learned the importance of adaptability, collaboration, and iterative problem-solving when building a complex product under time constraints.

What's next for Serasi

Dual MCP integration between MongoDB MCP and Google Calendar MCP: MongoDB MCP is live for booking persistence. Google Calendar MCP is the immediate next integration, enabling two-way sync so confirmed bookings automatically appear on both the customer's and hairstylist's calendars.
Payment integration: the platform flow is complete up to booking confirmation, but payment processing isn't yet connected. Next step is integrating a payment gateway so the end-to-end journey — consultation → booking → payment — is fully automated within Serasi.
Post-booking agent pipeline (immediate next), the unbuilt half of the system includes,
post_service_agent: collects reviews, sends follow-up messages, triggers loyalty points; and
product_agent: recommends haircare products used during the service for at-home purchase.