AI Freelancing Agent: Intelligent Job Discovery and Proposal Automation Case Study
AFA automates the most time-consuming part of freelancing: discovering jobs and crafting proposals. By combining job scraping, AI generation with Groq, and Slack notifications, it delivers curated opportunities directly to your inbox with ready-to-submit proposals, all tracked in a PostgreSQL audit trail.
Service
Automation
Industry
Freelancing / Productivity SaaS
Category
AI AUTOMATION
Built with: n8n, PostgreSQL, Groq, Slack.
Overview
The Problem
Freelancers spend the majority of their time on non-billable work such as scrolling job boards, filtering spam, and writing custom proposals for every opportunity instead of focusing on billable client work
Job postings expire within hours on competitive platforms, making manual job discovery workflows ineffective and causing missed high-value opportunities
Proposal quality varies based on energy levels and context switching, resulting in inconsistent rejection rates and reduced conversion rates
Freelancers lack a unified view of opportunities across multiple platforms and have no feedback mechanism to understand which proposals convert or why
The Solution
Automated job discovery pipeline that scrapes Freelancer, RemoteOK, and WeWorkRemotely in parallel with configurable filtering by skills, budget range, and job type
AI proposal generation powered by Groq that reads job descriptions, extracts requirements, and generates personalized proposals in under 10 seconds with consistent quality and tone
Slack-first notification workflow that delivers job details and AI-generated proposals to a review channel, allowing human approval before submission with direct links to original postings
PostgreSQL database that maintains a complete audit trail of discovered jobs, generated proposals, and submission status for analysis and optimization
Extensible n8n workflow architecture that enables adding new job boards, changing AI providers, and adjusting proposal templates without writing code
Solution Workflow
Discover
Generate
Notify
Track
Full Technical Breakdown
Job Discovery Pipeline (AFA 03)
Scheduled workflow that triggers daily and executes three parallel scraping branches against Freelancer, RemoteOK, and WeWorkRemotely using HTTP requests and DOM parsing. Results are deduplicated by job URL, stored with platform, company, title, budget range, description, and posting date. Duplicate detection prevents redundant processing of the same opportunity.
AI Proposal Generation (AFA 04)
Reads unprocessed jobs from PostgreSQL, constructs contextual prompts including job title, description, budget, and required skills, and sends requests to Groq API using Claude or Llama models. Captures generated proposals, stores them with model name, token count, and timestamp, and marks jobs as processed with status tracking.
Slack Notification and Routing
After proposal generation, formats and sends structured Slack messages to the review channel containing job details, full proposal text, direct platform links, and optional approval buttons. Uses ngrok to expose local n8n instance over HTTPS for webhook callbacks. Slack delivery failures don't block workflow execution; proposals remain stored for manual delivery.
PostgreSQL Data Model and Persistence
Maintains jobs table with job_id, platform, company, title, description, budget range, currency, URL, and timestamps; proposals table with proposal_id, job_id, proposal text, model name, token count, and status tracking. Implements database indexes on job_id, platform, and created_at for efficient filtering and aggregation queries.
Error Handling and Resilience
Failed scrapes log to PostgreSQL for manual review. Groq API timeouts and rate limits trigger exponential backoff retries. Slack delivery failures queue messages for manual resend. Database constraints prevent duplicate job insertion. Transactional integrity ensures proposal inserts and status updates maintain data consistency.
Extensibility Architecture
Job discovery accepts new HTTP nodes for additional job board platforms without schema changes. AI generation swaps between Groq, OpenAI, Anthropic, or local Ollama by changing API nodes. Proposal tone adjusts via Code node modifications. Future auto-submission capability planned for platforms with REST APIs like Freelancer.
Results
Key Features
Multi-Platform Job Aggregation: Discovers and deduplicates jobs from Freelancer, RemoteOK, and WeWorkRemotely in parallel using a single workflow. Eliminates manual checking of multiple job boards and ensures no opportunities are missed across platforms.
AI-Powered Proposal Generation: Generates personalized, contextually relevant proposals in under 10 seconds using Groq API. Maintains consistent tone, structure, and quality regardless of time of day or freelancer fatigue level.
Slack-First Review Workflow: Delivers proposals to a dedicated Slack channel with job details and direct links before submission. Keeps humans in control of high-stakes decisions while automating repetitive discovery and drafting work.
Complete Audit Trail: Logs every job discovered, every proposal generated, and every submission in PostgreSQL. Enables analysis of conversion rates, platform performance, and proposal effectiveness for continuous optimization.
Seamless Integration with Existing Workflows: Job links connect directly to original platforms for submission through native UIs. Works as a supplement to existing freelancing practices with no lock-in or workflow disruption.
Configurable Filtering and Parameterization: Job discovery filters by skill keywords, budget range, job type, and other criteria without code changes. Workflows use parameterized configuration, enabling customization for different freelancer profiles and preferences.
Before
Freelancers faced significant operational friction in their job discovery and proposal process. The typical workflow required manually checking multiple job boards each morning, often spending 30 to 60 minutes writing customized proposals for each opportunity. This manual process resulted in missed opportunities due to time constraints and posting expirations, with proposals frequently being rejected due to inconsistent quality or poor fit with job requirements. Freelancers had no unified view across platforms and lacked feedback mechanisms to understand which proposals converted or why, making it impossible to optimize their approach or identify patterns in high-performing submissions.
After
With AFA in place, freelancers now receive curated job opportunities automatically delivered to Slack, complete with AI-generated proposals ready for review. Job discovery and proposal generation are completed in under two minutes per job compared to the previous 45+ minute manual workflow. The system processes 300+ job listings daily across all platforms instead of the 20 to 30 manually reviewed opportunities. Proposal quality has become consistent through AI generation, eliminating variations caused by fatigue or context switching. Freelancers maintain complete control through a human-in-the-loop review process in Slack before submission, while building a comprehensive historical record that enables data-driven analysis of performance.
What This Proves
AFA demonstrates the practical application of AI in production automation, showing that integrating third-party LLMs into workflows is straightforward with proper API design and error handling. The system proves workflow orchestration capabilities by handling job scraping, AI generation, database writes, and Slack notifications across three platforms in parallel without manual intervention. The complete PostgreSQL audit trail of discovered jobs and generated proposals enables data-driven analysis of which platforms yield the best opportunities, which proposal tones convert effectively, and which skill areas command the highest rates. By automating the repetitive, low-skill work of discovery and drafting while preserving human control over high-stakes decisions like proposal submission, AFA illustrates a user-centric approach to automation that enhances rather than replaces human judgment.
Tech Stack
Data & Persistence
PostgreSQL — Self-hosted relational database storing jobs and proposals with full audit trail, indexes for fast filtering by platform and date, and transactional integrity
Docker Compose — Containerization and local deployment environment enabling reproducible setup with n8n, PostgreSQL, and optional Redis services
Orchestration Layer
n8n — Visual workflow automation platform running in Docker, handles job discovery, AI calls, database operations, and Slack notifications across 135+ configured nodes
AI & Language Models
Groq — Fast inference API for proposal generation using Claude 3.5 Sonnet, Llama 2, and Mixtral models with structured prompts and response logging
Notification & Messaging
Slack API — Real-time delivery of formatted proposal messages to review channels with Block Kit formatting, direct platform links, and optional approval workflow
ngrok — Public HTTPS tunnel for webhook callbacks and local development webhook exposure
Scripting & Data Handling
JavaScript — n8n Code node scripting for data transformation, job URL deduplication logic, proposal formatting, error handling, and retry logic
Scalability & Flexibility
Add New Job Platforms: Extending to Upwork, PeoplePerHour, or other platforms requires only a new HTTP scraping node in the AFA 03 workflow with no database schema modifications.
Swap AI Providers: Changing from Groq to OpenAI, Anthropic Claude, or self-hosted Ollama requires modifying a single API integration node, preserving all prompt logic and data handling.
Adjust Proposal Tone and Style: Customizing proposal voice, formality level, and structure is achieved through Code node prompt modifications without rebuilding workflows or affecting data pipelines.
Implement Auto-Submission: Platform APIs like Freelancer's can be integrated into a new AFA 05 workflow to automatically submit proposals, building on existing job and proposal data structures.
Support Multiple Skill Profiles: Creating separate discovery workflows for different skill sets or experience levels requires only parameterized filter changes, enabling operation across diverse freelancer personas.
Expand Budget and Currency Handling: The PostgreSQL schema supports multiple currencies and complex budget logic. Workflows can be enhanced to support hourly, fixed, and retainer-based project types through configuration.
Real-Time Notifications and Webhooks: Extending Slack notifications to Discord, email, SMS, or push notifications requires adding output nodes to existing workflows without changing core discovery or generation logic.
Rate Limiting and Queue Management: Redis integration enables sophisticated queuing, rate limiting across APIs, and request prioritization to handle scaling to thousands of daily job discoveries.
Historical Data Analysis: Complete PostgreSQL audit trail enables future ML models to predict proposal conversion rates, identify high-performing job descriptions, and optimize freelancer positioning based on historical outcomes.