Helix

Overview

At Helix I worked on the member platform and the AI workflows around community calls. The work covered Google Drive ingestion, LLM extraction into PostgreSQL, the Next.js member experience, retrieval-augmented chat, and scheduled agents. Details here stay at the level of architecture and contribution — no internal product data.

Problem

Community knowledge lived in call recordings and scattered member context. The team needed structured records, a member-facing product, and assistants that could retrieve that context without leaking the wrong data.

What I built

An ingestion pipeline pulled recordings from Google Drive and used LLMs to extract member wins, commitments, and profile facts into PostgreSQL. The member app covered directory, profiles, onboarding, and invites, with row-level database rules. A RAG assistant searched transcripts and member records with embedding search and per-request token cost tracking. Scheduled agents produced weekly summaries and introduction suggestions.

My contribution

• Built the Google Drive → LLM extraction → PostgreSQL ingestion pipeline.
• Developed directory, profiles, onboarding intake, and invite flow in Next.js.
• Shipped retrieval-augmented chat with embedding search and token cost tracking.
• Delivered scheduled agents for weekly summaries and introduction suggestions.

Technologies

• Next.js
• PostgreSQL
• LLMs
• RAG
• Embeddings
• AI agents