career-agent

Design principles

• Human-centric
  • In most cases, AI should augument, not replace the human from tasks
• AI Collaboration
  • The quality of the end result is the product of human-AI collaboration
• Transparency
  • The user should be able to see and edit the data in all it's intermediary forms

TODO: High Level

• Prototype the main processing steps (see diagram.excalidraw and Project Description section) in jupyter notebooks
• Build an evaluation set for each step of processing that uses LLMs
• Decide which platform to release on

Setup

• Get an API key from OpenAI. Paste it into a file called .env in the root directory of this repo.
  • Like this: OPENAI_API_KEY=<the key>

Python

• Make sure the python version is 3.8
  • Check with: python --version
  • If it's not 3.8, install pyenv and then run:
    • pyenv install 3.8
  • N.B. don't use the Terminal window in VSCode, it doesn't work well with pyenv. Use the terminal app.
• Create a virtual environment
  • python -m venv venv
• Install the requirements
  • pip install -r requirements.txt

Optional

• Readme files will contain wikilinks to other files in the repo. To use these links, use an extension e.g.
  • Foam for VSCode
  • ... or Markdown Links extension in VSCode.
• View Excalidraw Diagrams
  • If using VSCode, install Excalidraw plugin

Project Description

1. Parse CVs
   • Create database of all your experiences, skills, deliverables etc.
   • Columns: Company, Experience, Skills, KPIs
2. Parse Job Specs and extract:
   • Technical job requirements
   • Soft skill requirements
   • Industry, culture and values from job spec
3. Match Experience to Job Spec
4. Write CV
   • For each work experience section, write the bullet points for key experiences
   • For profile show it the:
     • Outputs of step 3.1 (the work experiences)
     • The name, values, culture, industry of the company
     • An conceptual outline / structure
5. Cover Letter
   • Body paragraphs with inputs. Reverse chronological order
     • CV work experience
     • Soft skills requirements
   • Intro paragraph summarising journey in chronological order, building a narrative for why the next logical step is in this role.
     • Body paragraphs from 4.1
     • Company details (industry, values, culture etc)
   • Summary paragraph
     • The rest of the cover letter
   • (Style matching)
     • In order to avoid looking like it was AI-generated, it would be good to provide examples of your own prose as part of the prompts.

Implementation Discussion

UI

Requirements

• Present intermediary data and make it editable before proceeding to next steps

Platform options

This is linked to choice of language. We'd prefer to use Python for the better libraries but TS/JS is better for web apps.

• Web app
  • When running locally, run on localhost
  • Could be hosted on the web
• Desktop app
  • Electron
• CLI
  • Could be a CLI tool
• Jupyter Notebook
  • Could develop like this and convert to a another platform later
• Extension to a text editor
  • VSCode
  • Obsidian
• Streamlit