AI Commodity Research & Trading Platform
Budget / Salary$30–250
TypeFreelance project
LocationRemote
Posted2 hours ago
I have a client who wants to create an personal use AI Commodity Intelligence Platform that:
Monitors weather, satellite imagery, government reports, and news.
Scores each commodity based on supply and demand conditions.
Identifies unusual shifts before they're widely recognized.
Produces a ranked watchlist with supporting evidence.
He has very limited personal budget, however, based on his GPT research, AI can realistically generate roughly 80-90% of the initial software code automatically.
GPT OVERALL RECOMENDSATIONS AND WORKFLOW:
Platform is NOT initially intended to trade automatically. Its job is to continuously:
Collect → Analyze → Compare → Score → Challenge → Recommend → Risk-check → Present
..and then client approves or rejects the proposed trade.
Need to Keep the infrastructure extremely simple. GPT recommends not to buy or configure a dedicated server yet due to small budget constrains.
GPT recommendations to start with:
Component Phase-1 choice
Development GitHub repository
Development environment GitHub Codespaces
AI programmer OpenAI Codex
Language Python
Database SQLite initially
Backend Python services
Dashboard Streamlit or similarly lightweight Python UI
Scheduling Python scheduler / cron-style job
AI analysis OpenAI API
Secrets Environment variables / GitHub secrets
Deployment Development cloud environment initially
Trading None initially—research only
GitHub currently includes a monthly Codespaces allowance with personal accounts, which can eliminate much of the local Python/environment setup your assistant would otherwise have to perform.
Codex is particularly appropriate here because OpenAI currently positions it as an agentic software engineer that can build features, perform refactors, fix bugs, review code and work across multiple engineering tasks.
That means developer doesn't need to learn enough Python to write the application. He just needs to learn enough to:
give instructions → run tests → inspect results → report problems → let Codex fix them.
Monitors weather, satellite imagery, government reports, and news.
Scores each commodity based on supply and demand conditions.
Identifies unusual shifts before they're widely recognized.
Produces a ranked watchlist with supporting evidence.
He has very limited personal budget, however, based on his GPT research, AI can realistically generate roughly 80-90% of the initial software code automatically.
GPT OVERALL RECOMENDSATIONS AND WORKFLOW:
Platform is NOT initially intended to trade automatically. Its job is to continuously:
Collect → Analyze → Compare → Score → Challenge → Recommend → Risk-check → Present
..and then client approves or rejects the proposed trade.
Need to Keep the infrastructure extremely simple. GPT recommends not to buy or configure a dedicated server yet due to small budget constrains.
GPT recommendations to start with:
Component Phase-1 choice
Development GitHub repository
Development environment GitHub Codespaces
AI programmer OpenAI Codex
Language Python
Database SQLite initially
Backend Python services
Dashboard Streamlit or similarly lightweight Python UI
Scheduling Python scheduler / cron-style job
AI analysis OpenAI API
Secrets Environment variables / GitHub secrets
Deployment Development cloud environment initially
Trading None initially—research only
GitHub currently includes a monthly Codespaces allowance with personal accounts, which can eliminate much of the local Python/environment setup your assistant would otherwise have to perform.
Codex is particularly appropriate here because OpenAI currently positions it as an agentic software engineer that can build features, perform refactors, fix bugs, review code and work across multiple engineering tasks.
That means developer doesn't need to learn enough Python to write the application. He just needs to learn enough to:
give instructions → run tests → inspect results → report problems → let Codex fix them.
Apply on Freelancer →
Project sourced from Freelancer.com. Applications happen directly on the original platform — we never collect your data.