App Development with Supabase and AI Email Parsing

I have an app (Supabase backend, hosted on Vercel) that lets users sign up for price alerts on products. Users can browse products, subscribe to alerts, and get notified when prices drop. I already have Resend set up for transactional email, and I’ve built an admin tool where I can manually search products, filter by time since last send, and trigger emails with properly formatted affiliate links and attribution.

The problem: the price monitoring and email sending is entirely manual. I need to automate it.

What Needs to Be Built

• 1. Confirmation Email on Signup
  When a user signs up for a price alert on a product, they should immediately receive a confirmation email ("You’re signed up for alerts on [Product Name]"). This doesn’t exist yet. Should use the existing Resend integration.
• 2. Automated Price Drop Detection via Email Monitoring
  Rather than integrating with a single price-tracking API, the system monitors a dedicated email inbox that’s subscribed to price alert services (CamelCamelCamel, Keepa, retailer newsletters, etc.). Every 'creator' account on dip would get their own dedicated email to subscribe to price alerts with.
  How it works:
  • A Vercel cron job runs every 30 minutes
  • Fetches unread emails from the monitoring inbox
  • Passes each email body to an LLM (Claude Haiku or similar) to extract: product name, ASIN, deal price, original price, retailer, deal URL, and a confidence score
  • Matches extracted product against the Supabase product database (ASIN match → URL match → fuzzy name match)
  • If confidence is high (≥0.85): auto-sends price alert emails to all subscribers for that product
  • If confidence is medium: flags for my manual review
  • If low: logs and discards
• 3. Price Alert Email Dispatch
  When a qualifying price drop is detected, send emails to all subscribed users for that product. The email sending and affiliate link generation logic already exists in my admin tool — this just needs to be triggered automatically instead of manually.

Existing Infrastructure

• Framework: on Vercel (serverless)
• Database: Supabase (Postgres)
• Email: Resend (already configured and working)
• Affiliate links: Already have a working system for generating properly attributed links
• Admin tool: Existing UI for manually selecting products, filtering by last send date, and triggering alert emails — the automation should reuse this logic, not rebuild it

Key Constraints

• Must run within Vercel's serverless environment (cron jobs, edge/serverless functions)
• Vercel cron has a max execution time — the inbox scan + LLM calls need to complete within that window or be chunked
• 24-hour cooldown per product to avoid duplicate alerts from multiple price services
• Minimum discount threshold (e.g., ≥10% or ≥$5) to avoid spamming users with trivial drops
• Minimum time since last alert threshold to avoid spamming users accidentally (I already have this built as a manual filter)

Deliverables

• Confirmation email flow wired into the existing signup process
• Email ingestion service (Vercel cron + IMAP integration)
• LLM parsing function that extracts structured deal data from raw emails
• Product matching logic against Supabase
• Auto-dispatch that triggers the existing email send logic for qualifying matches
• Logging and a simple way for me to review flagged (medium-confidence) matches

Open Questions for Discussion

• Best approach for email inbox access within Vercel serverless
• Whether Vercel cron execution limits require batching the inbox scan
• Your recommendation on LLM provider/model for the parsing step (cost vs. accuracy)

Contract duration of 1 to 3 months, with 30 hours per week.

Mandatory skills: JavaScript, React, HTML, Email Deliverability, Supabase, API, Data Scraping, Vercel