StyleSeat
Beauty & wellness marketplace · via Revelo
Booking2026
Senior Software Engineer · via Revelo
Context
StyleSeat is a marketplace where beauty and wellness professionals publish their availability and clients book directly. The booking flow is the centre of the product: if scheduling is hard, the professional loses income and the client goes elsewhere.
I joined through Revelo as a dedicated senior engineer, working embedded with their product and engineering teams rather than as an external vendor.
The challenge
A stylist loses hours every day confirming appointments, chasing no-shows and rescheduling, and that time is unpaid. Automating it with a model means solving the hard parts first: the agent must not invent availability, must not do anything irreversible without permission, and must leave a record of everything it did.
What it does
- Agents that confirm appointments and reschedule for the professional
- Schema-defined tools with confirmation on irreversible actions
- Versioned workflows, staging and rollback
- Professional availability and scheduling
- Client-facing booking flow (web and mobile)
- Lifecycle messaging and unified event tracking
What I did
- Built the agent-driven workflows: the agent reads the calendar and booking state, drafts client messages and applies booking changes inside the limits the professional sets.
- Mine end to end were the tools the model can call, their schemas, and the orchestration around them: retries, timeouts and a deterministic fallback for when the model is down.
- Put the guardrails in place: availability always comes from a real call to the scheduling service and never from model text, irreversible actions require confirmation, and every action is logged with what the agent saw and what it called.
- Operated the workflows as well as building them: versioning with rollback, staging validation against real booking scenarios, role-separated access, and token budgets, rate limits and quotas per workflow for peak hours.
- Integrated Braze for lifecycle messaging and Segment as the tracking layer, so the agent was measured like any other feature of the product.
- Established and optimized the booking features: availability and scheduling, reservation flows and the client-facing experience on web and mobile.
Stack
- Python
- DjangoCelerypytest
- Ruby on Rails
- ActiveRecordActiveJobSidekiqRSpec
- JavaScript
- ReactReact NativeTypeScript
- Data
- PostgreSQLRedis
- Product and user data
- BrazeSegment
- AI
- AI agentsTool callingLLM APIsEvalsGuardrails
Services practised here
- Backend Architecture & APIsYour product ends up with a backend that survives growth: services with clear boundaries, documented APIs and a data model built for the queries that actually run.
- AI Integration & AutomationModels and agents go into the product you already have with the same discipline as any other service: a contract of your own, cost limits and a log of what they did.
Outcome
Two and a half years on the same product: the work moved from shipping features to sustaining and evolving the core of the business.
With agents, what decides whether the thing works is not the model. It is where the data comes from, what the tool can and cannot do, and what happens the day the model is down.