Portfolio LLM

Inspiration

Wanted a personable “digital twin” that lets recruiters explore Kevin’s work the way they would in a real conversation.

What it does

Streamlit app with a recruiter-facing chat powered by Claude and a self-service admin panel. It ingests resumes, project notes, and quick updates into Supermemory so answers stay current

How we built it

Streamlit for the UI (chat, profile, admin tools) Supermemory v3 REST API for document upload/search Anthropic Claude for retrieval-augmented responses Session state + sidebar actions to manage uploads, manual notes, and model selection

Challenges we ran into

Getting Supermemory endpoints right (v3 vs v4) and making uploads work with existing containers Handling model availability errors (404s) across different Claude tiers Balancing conversational tone with concise, recruiter-friendly answer

Accomplishments that we're proud of

Toggleable profile summary and quick notes make it easy to keep content fresh Clean, non-robotic replies that still stay grounded in verified documents

What we learned

Supermemory requires precise API usage; small payload differences can cause 404s Giving the user control over model selection is vital when feature access varies by account Streamlit’s session state is perfect for instant UI updates (like show/hide summary)

What's next for Portfolio LLM

Build richer document organizers (tagging, version history) Add GitHub auto-fetch back in once Supermemory containers scale up Layer in interview-style prompts so the AI can proactively highlight standout projects

Built With

• python
• streamlit