Saumya Chaturvedi and I were invited by Cecile Tamura and Kevin Filson to talk about our LoRACode paper on Ploutos AI.
We walked through the core idea: using LoRA adapters to fine-tune code embeddings for semantic search, training less than two percent of the model parameters while staying competitive with full fine-tuning. We also got into some good discussion about where this fits in the broader code retrieval landscape.
The recording is up if you want to watch: Ploutos AI stream.