AI-Native Vertical SaaSActive product

Placement OSInception to Offer

My roleProduct and systems architect

ContextMBA candidates, mentors, institutions

System classAI-Native Vertical SaaS

FocusCandidate state, workflow, authorization

Brief

The problem

MBA candidates moved between disconnected CV tools, courses, mentors, practice cases, and application trackers. No shared record connected a role target, evidence gaps, practice history, and next action.

Open product preview

Constraints

One journey, many roles

01The experience needed to support first-year internships, final placements, mentors, and administrators.

02The browser should remain thin; core state, validation, and authorization belong on the server.

03The product needed role-specific journeys across marketing, business analysis, product, and case interview tracks.

04Multi-tenancy, policy enforcement, and observability were required from the platform foundation.

Architecture

The candidate record

A long-lived candidate record connects diagnostics, practice, mentoring, applications, and next actions.

Placement OS / logical architectureSanitized view

01ExperienceGo templ + HTMX + Alpine

02ApplicationEcho modular monolith

03Domain modulesResumes, interviews, applications, diagnostics

04PlatformPostgreSQL, Redis, Casbin

05OperationsOTel, rate limits, CSRF, tenant context

Decisions

Why the monolith came first

A modular monolith before distributed services

Strong module boundaries preserve speed and consistency while leaving extraction paths for future scale.

Server-driven UX for workflow-heavy screens

Go owns validation, identity, search, and pagination; HTMX progressively updates the interface.

Candidate state is the product spine

Diagnostics, CV evidence, practice, mentoring, and applications update one longitudinal career model.

AI is embedded at decision points

Agents diagnose gaps, improve evidence, run practice, and interpret progress at decision points.

Production

Controls across the journey

01Casbin-backed authorization and tenant-aware middleware.

02CSRF, identity, organization, and rate-limit layers.

03PostgreSQL as system of record with Redis for ephemeral/queued concerns.

04OpenTelemetry across request and workflow boundaries.

05Modular domain ownership for maintainability and team scale.

GoEchotemplHTMXAlpine.jsPostgreSQLRedisCasbinOpenTelemetryModular monolithMulti-tenancyRBACAI UXDomain modeling

Outcome

One record connects role targets, practice, mentoring, applications, and offer preparation.

Relevant problem?

Modeling a multi-role product around one durable journey?

Discuss vertical AI productsAll work

Next case study