GhoreKaj — Designing Trust Infrastructure for Bangladesh's Domestic Worker Economy
- 3customer/worker/admin
- Role portals
- 12lifecycle steps
- Booking states
- 6tested paths
- Core flows
Project Overview
GhoreKaj is a full-stack domestic service marketplace I designed, engineered, and shipped for the Bangladesh market. The platform connects urban households with verified local domestic workers and supports the full service lifecycle from task posting to bid review, job confirmation, completion, dispute handling, and payment reconciliation.
The project was built as a solo product effort using an agentic AI workflow. I directed the product strategy, UX architecture, UI design, implementation priorities, testing, debugging, and deployment while using AI as a collaborator for acceleration.
This case study is public-safe: sensitive implementation details, credentials, private routes, private API details, exact security internals, and vulnerable operational details are intentionally omitted or generalized.
Design note
The product challenge was not just matching supply and demand. It was designing enough trust infrastructure for strangers to complete a high-trust household service transaction.
The Trust Gap
Bangladesh's domestic labor sector still operates heavily through informal word-of-mouth channels. Customers often cannot verify who they are inviting into their homes, and workers have no portable reputation when moving between jobs.
The core problem was a bilateral trust deficit. Customers need identity confidence, service history, and accountability. Workers need fair access to clients, proof of reliability, and a way to build reputation beyond personal referrals.
The design challenge was to create structure without overwhelming low-digital-literacy users or forcing Western marketplace assumptions into a local, cash-first service culture.
Goals and Success Criteria
The MVP needed to support three complete role experiences: customers posting tasks and hiring workers, workers browsing open work and submitting bids, and admins verifying identities and resolving operational issues.
Success meant a customer could post a task, receive bids, accept a worker, confirm service with OTP, and complete the job flow. Workers needed a clear verification path, task feed, bid flow, and job status tracking. Admins needed queues for verification, tasks, payments, disputes, and reviews.
The product also had to feel credible on mobile, avoid blank first-login states, and communicate trust through visible verification, status labels, and guided transitions.
Market and User Context
The research direction focused on household service hiring behavior in Bangladesh: referral networks, community trust, mobile money familiarity, shared Android devices, and the practical importance of identity signals.
Three primary users shaped the experience: urban household customers seeking safe help, domestic workers seeking consistent work and fair pay, and platform operators responsible for verification and dispute handling.
A key insight was that users do not want an abstract marketplace. They want a structured version of a familiar local behavior: someone trusted, nearby, available, and accountable.
Design and Engineering Process
I started by mapping the service lifecycle as a state machine, because every payment, notification, dispute, and button state depends on the current job status. This reduced ambiguity and prevented role-specific actions from appearing at the wrong time.
Task creation was redesigned as a progressive wizard instead of a long form. Location, service category, schedule, budget, and review steps were separated so users could focus on one decision at a time.
The worker and admin workflows were built around queues: available tasks, submitted bids, accepted work, verification requests, payment reviews, and disputes. This made the product operational rather than just visually complete.
Design note
Agentic AI accelerated implementation and debugging, but I owned the requirements, UX decisions, state model, review, and final quality gate.
The Marketplace System
The final product includes customer, worker, and admin portals with role-aware navigation and status-specific actions. Customers can post tasks, compare bids, shortlist workers, accept a bid, track active work, and manage completion.
Workers can register, complete verification, browse available tasks, submit bids, receive status notifications, and track work progress. Admins can review identity documents, monitor marketplace activity, manage categories, resolve disputes, and reconcile payment references.
Trust is expressed through practical product mechanics: verification gates, OTP job confirmation, visible worker profiles, bid context, and auditable manual payment records.
Accessibility and Inclusive UX
Accessibility decisions focused on readable contrast, mobile-first form behavior, clear labels, visible status states, and straightforward language. The product needed to work for users with different literacy levels and device constraints.
I avoided relying only on color for status communication and used explicit labels for task states, verification status, payment status, and dispute status.
Future validation should include keyboard walkthroughs, screen-reader checks, and direct testing on lower-end Android devices.
Outcomes
GhoreKaj became a production-deployed product that demonstrates marketplace UX, multi-role system design, trust mechanics, and full-stack implementation depth.
The project shows my ability to turn a socially complex, high-trust service problem into a structured digital product with coherent roles, flows, and operational controls.
Recommended future metrics include task posting completion rate, bid acceptance rate, worker verification time, job completion rate, payment reconciliation time, and dispute frequency.
What I Learned
This project reinforced that local context matters more than copying familiar marketplace patterns. For this market, trust, verification, payment culture, and service confirmation are core UX concerns.
It also clarified how powerful agentic AI can be when paired with strong product ownership. AI helped move faster, but the quality came from clear direction, testing, review, and local product judgment.