Inspiration
The gig economy powers millions of workers, but trust remains its biggest friction point. Cleaners face inconsistent work and unclear protections. Clients must trust strangers entering their home based on profile photos and star ratings. PureTask Trust asks: What if trust could be built into the workflow itself?
For real-world services—cleaning, home care, maintenance, inspections—trust requires more than ratings. It requires clear scope, verified identity, transparent expectations, evidence of completed work, fair dispute resolution, and intelligent systems that help both sides make better decisions.
Gemini became the intelligence layer behind this vision, helping us build systems for job matching, evidence analysis, service quality validation, and contextual decision-making throughout the professional and customer journey.
What We Built
Core Platform
- Marketplace for service professionals (launched in beta with test data)
- Complete job lifecycle: request → matching → in-progress → evidence → payment → reputation
- Mobile-first cleaner workflow: clock in → verify scope → capture before/after → checklist → clock out
- Integrated payments (Stripe), identity verification (Checkr), SMS (Twilio), email
- Evidence pairing: before/after photos tied to specific rooms/tasks, not generic galleries
Gemini-Powered Features (Deployed)
- Intelligent matching: Analyzes cleaner profiles and job requirements to optimize assignments
- Evidence analysis: Evaluates before/after photo pairs for consistency, lighting, coverage
- Scope interpretation: Understands customer service requests and extracts structured task requirements
- Quality signals: Detects missing or inconsistent evidence, flags potential issues
- Policy reasoning: Applies cancellation ladder, review windows, and dispute rules in context
Architecture
Built on Vite + React 19 + Supabase PostgreSQL. The platform uses a job state machine (requested → matched → accepted → in-progress → evidence-submitted → completed/disputed) as a reliable source of truth. Gemini integrations are separated by reasoning task (matching, evidence, scope, quality) rather than one generic prompt.
How It Works
For Cleaners
- Browse matched jobs (Gemini considers availability, service area, ratings, job complexity)
- Accept and clock in
- Review scope (AI-extracted task checklist)
- Capture before/after evidence for each room
- Complete verification checklist
- Clock out and submit
For Customers
- Describe cleaning request (Gemini structures the scope)
- Review matched cleaner profile and history
- Confirm booking and payment hold
- View before/after evidence in real time
- Approve completion (or dispute with context)
For Disputes
Gemini assembles scope, messages, job timeline, evidence, and policy into a structured case for human review—AI informs, humans decide.
Technical Challenges We Solved
Trust is not one feature—it's a system. Identity alone doesn't prove work quality. Photos without context don't prove scope completion. We designed multiple signals (verified ID + job state + evidence + timeline + payment) to work together.
AI should assist judgment, not replace it. Real-world service work is messy. Lighting changes. Rooms move. Customers add scope. We designed Gemini outputs as trust signals (confidence + reasoning) not verdicts. Humans make final calls on disputes.
Offline resilience. Cleaners work in basements and buildings with dead zones. We persisted timers, job state, and evidence locally so losing connection doesn't lose work.
Balancing accountability with usability. If evidence collection is too complex, workers hate it. If too light, it loses value. We landed on: one before/after per room, structured checklist, optional detailed notes.
Bidirectional protection. Customers need protection from incomplete work. Professionals need protection from vague scopes and false claims. We built guardrails for both.
What's Next
- Deeper multimodal evidence analysis (Gemini understanding room condition changes over time)
- Stronger dispute resolution UI (assembling evidence, scope, messages into review cases)
- Recurring service coordination (Gemini learning repeat customer preferences)
- Professional business tools (estimates, route optimization, earnings analytics)
Longer term: the same trust infrastructure can extend to home care, inspections, maintenance, landscaping, and other industries where strangers coordinate physical work.
Key Differentiator
PureTask doesn't use Gemini as a chatbot overlay. It's woven into core workflows—matching, evidence review, scope interpretation, quality validation—making trustworthy behavior easier to perform, easier to verify, and easier to reward.
Log in or sign up for Devpost to join the conversation.