Back to selected work

Project summary

AI & Automation
Internal application

RB Trends Internal Operations Platform

An internal tool that ranks Upwork jobs by suitability, matches each brief with relevant portfolio work, and helps prepare tailored proposals.

From job listing to considered proposal

Evidence and decisions

problem
Finding suitable work takes more than keyword search; each opportunity needs a fit check, relevant evidence, and a proposal tailored to its brief.
approach
Combined job ingestion and filtering with AI-assisted analysis, portfolio matching, and proposal preparation in a repeatable Django workflow.
value
Gives proposal work a consistent path from opportunity review through evidence selection and drafting.
Evidence boundary
A committed job-ranking and hours-estimator snapshot is available from 27 November 2025. It is archival evidence; no accuracy or revenue uplift is claimed.

Overview

This internal research and proposal tool helps turn an Upwork job listing into a considered proposal. It ranks opportunities against skills and budget, finds relevant portfolio evidence, and supports structured, AI-assisted drafting.

The available screenshot is an archived repository snapshot from November 2025. It does not establish proposal accuracy or revenue uplift.

Delivery surface

  • Opportunity filtering by skills, budget, and suitability
  • Portfolio evidence matching against job requirements
  • AI-assisted proposal preparation
  • Ranked job and hours-estimator research snapshots

Outputs and artefacts

Ranked opportunity reportsRelevant portfolio evidence for proposal briefsStructured proposal draftsHours-estimator snapshots

Technology stack

PythonDjangoOpenAIscikit-learn

Publication boundary

This page describes the system at a useful level without exposing client names, private records, credentials, or unverified commercial outcomes.

Next step

Want the build, not just the summary?

If this is the kind of system you need, the next conversation is usually about the workflow bottleneck, the smallest useful first version, and where the review loop needs to stay visible.