Data Intern (Gen AI)
Seaspan ULC
Built an AI agent with LangChain and Pinecone to index 691 technical documents and retrieve cited results.
What I Did
I built an AI agent using LangChain and Pinecone that indexed 691 unstructured technical documents, extracted structured fields from free-text records, and returned cited results through a query endpoint. I evaluated retrieval precision on test sets, logged quality drift over time, and flagged stale records for re-indexing.
Impact
The agent gave the operations team fast retrieval over 691 documents with cited answers. The quality monitoring caught retrieval degradation early and kept the index fresh.
What I Learned
I learned to build retrieval pipelines with LangChain and Pinecone, extract structured fields from unstructured text, and monitor retrieval precision over time to detect quality drift.
Key Highlights
Built an AI agent with LangChain and Pinecone that indexed 691 unstructured technical documents, extracted structured fields from free-text records, and returned cited results via a query endpoint.
Evaluated retrieval precision on test sets, logged quality drift over time, and flagged stale records for re-indexing.