Mailflow — AI-Assisted Job Outreach
An AI-assisted application that turns job descriptions and candidate context into concise, tailored cold emails. It combines secure user accounts, CV text extraction, reusable templates, and direct email delivery in one workflow.
AI-assisted job outreach with CV text extraction, reusable templates, secure user data, and email delivery
A secure, user-centered workflow that grounds concise outreach emails in a job description and candidate-provided context, with review and delivery controls built in.
Mailflow streamlines personalized job outreach while keeping candidates in control of every message. Users provide a job description, recipient details, and their professional context; the application creates a concise, role-specific email that can be reviewed, copied with formatting preserved, or sent directly.
Core workflow
- ▸Secure email/password authentication with Supabase Auth
- ▸Profile settings for portfolio, LinkedIn, reply-to address, signature, and email appearance
- ▸Job-description, recipient, language, and candidate-context inputs in a focused compose flow
- ▸PDF, DOCX, and TXT CV text extraction (up to 5 MB) for generation context
- ▸Reusable prompt templates with selectable tone labels
- ▸Rich HTML and plain-text copying, plus direct sending and searchable history
AI and delivery pipeline
Candidate context + job description + selected template
→ authenticated Next.js API route
→ Gemini generates a subject line and concise HTML email
→ styled preview, copy-to-clipboard, or Resend delivery
→ Supabase stores the email history and delivery status
Technical design
- ▸Next.js 16 and React 19 application written in TypeScript
- ▸Supabase Auth and PostgreSQL with Row Level Security to isolate user profiles, templates, and email history
- ▸Gemini-powered generation instructed to avoid inventing qualifications
- ▸Resend delivery with persisted draft, sent, and failed statuses
- ▸Server-side document extraction using pdf-parse and mammoth
Responsible use
- ▸Generated emails are starting points; users review claims, links, names, and recipients before sending
- ▸CV uploads are parsed for context and are not persisted by the current flow
- ▸Sending is designed for legitimate, personalized outreach in line with anti-spam and privacy requirements.