Your job searcher

Your Job Searcher is not just an agent; it’s a new way of working. It performs live, autonomous job searches on your favorite platforms like Google and LinkedIn, so you don't have to.

Inspiration: people searching for jobs every single day—manually making cover letters, sending emails, and just getting completely exhausted. I wanted to build a system that takes that burden away and gives people their time back.

What it does: Your Job Searcher automates the entire application pipeline using a coordinated team of AI workers:

Agent 1 (The Scout): It checks your real-time personal information and searches for live jobs on Google and LinkedIn that perfectly match your data, focus, and skills.

Agent 2 (The Pitchman): You no longer have to be exhausted every day. Pitchman takes over the heavy lifting by writing a highly targeted cover letter based on your real-time data using Gemini 3.5 Flash.

Agent 3 (The Closer): Sending emails is boring. Hand that work over to Agent 3! It easily drafts and pushes live emails to your Gmail in seconds using the latest Gemini 3.5 Flash.

Agent 4 (The Reporter): No, no, no!!! You don't even have to manually check up on every agent. Agent 4 monitors the entire process and writes a final execution report for you.

How we built it:

Core Logic: Python, orchestrated via a streamlined multi-agent pipeline.

Frontend & Hosting: Streamlit, containerized via Docker, and deployed live to Google Cloud Run.

AI Models: Gemini 3.5 Flash (via the new google-genai Vertex AI SDK) and Gemma 270M.

Cloud Infrastructure: Heavy reliance on Google Cloud Platform, including Cloud Storage, Compute Engine (e2-micro hosting Gemma), and Secret Manager to securely handle API keys and OAuth tokens in production.

Challenges we ran into: Adapting to the new Automatic Function Calling (AFC) requirements in the absolute latest Google Gen AI SDK took some debugging. Another major challenge was finding the best, most cost-effective way to host Gemma locally on the cloud. I finally got it sitting safely and running smoothly inside the "Always Free" tier of a GCP e2-micro VM!

Accomplishments that we're proud of: Building a system that isn't just wasting money on something useless. I built a web app on GCP before, but I honestly didn't feel great about it because Gemini Notebook (NotebookLM) was already doing the job better. But for this agentic AI system? I personally feel that it is incredibly useful, highly practical, and I am genuinely happy with the result.

What we learned: The most important thing I learned is that AI is great, but you still have to look after it. Even if you use AI as a partner, you have to read its answers, review its results, and sometimes remind it of what you were doing to keep it on track. The second biggest thing I learned was how to use Google Cloud VMs (GCP Compute Engine).