Work in progress...
The Talent Operating System, an end to end recruitment engine from a single source, a job description.
The prototype, built using AWS PartyRock, a no code widget drag and drop AI app builder: https://partyrock.aws/u/dregan86/bGmkKl-1F/Talent_Operator_OS
The goal, to generate a recruiting guide for talent operators with further improvements planned, such as increased automation and autonomous reasoning.
The TalentOS creates a recruiting playbook, from role understanding, talent insights, research, sourcing reconnaissance, sourcing strategy, recruitment marketing, candidate evaluation, offer strategy, and recruitment analytics
I will attempt to build this system out the below architecture to power a real version of the modules shown in the PartyRock app.
- Inference Layer
Engine: Ollama (Windows) + OLLAMA_VULKAN=1. Primary: Qwen 2.5 Coder 7B (5-bit) – ~5GB VRAM. Vision: Qwen 2.5-VL 3B – ~2GB VRAM.
- Interaction Layer
Open WebUI (Docker). Persistent Memory: Mem0 + ChromaDB (Docker).
- Agentic Layer
VS Code + Roo Code extension. Aider (CLI).
- Intelligence Pipeline
Python via uv. PyGithub + SQLite + pgeocode.
- Visualization & BI
Kepler.gl (GPU-accelerated). Evidence.dev.
- Infrastructure
Tailscale + Docker Desktop. Firewalla Purple + UniFi monitoring.
Update: 03/09/2026
In about a week's time, mostly in the evenings, I created this website from the github.com/marpeand/lotse template while leaning into AI and vibe coding. It's been an incredibly fun learning experience, exploring tech ecosystems, and working through break fix roadblocks along the way.
I've learned how to navigate GitHub, discovered website templates, deployed using Vercel, mostly copied and pasted PowerShell and Git Bash scripts, deployed a Docker container, worked with SVG image files, used VS Code, vibe‑coded with v0 | Claude | GitHub Copilot, got introduced to persistent memory, started to understand GPU/CPU performance and limitations, and installed and configured LLMs.
This might be old news for some, but it was all brand new to me, and I can't wait to start using this system.
