Enzo Incutti
Building code review agents at Recurse
Current Focus
I'm currently building code review agents at Recurse, where I work on infrastructure that helps catch bugs before they ship. I am particularly interested in code models, specifically code review models - think of knowing where a critical bug is introduced the moment you type the wrong keystrokes.
Technical Systems
Things I helped build from the ground up
Agent Infrastructure
- • Core agent architecture with context management, tool calling, structured outputs, and observability (h/t Requesty)
- • Evaluation pipelines integrated with Weights & Biases for continuous model improvement
- • Elasticsearch-powered knowledge base with vector embeddings for semantic code search and documentation integration
Production Systems
- • CI/CD pipelines on GitHub Actions, GCP CloudBuild, and CloudRun
- • GitHub App with FastAPI webhooks for real-time PR monitoring and review
- • Redis-based async processing architecture for scalable background analysis
- • Terminal UI with click that brings code review capabilities directly into the development workflow—catching issues before they reach PR stage
Research That Shapes My Work
- On Usefulness of the Deep-Learning-Based Bug Localization Models to Practitioners
- RepoBench: Benchmarking Repository-Level Code Auto-Completion Systems
- Alibaba LingmaAgent: Improving Automated Issue Resolution via Comprehensive Repository Exploration
- CoRNStack: High-Quality Contrastive Data for Better Code Retrieval and Reranking
- CodeT5: Identifier-aware Unified Pre-trained Encoder-Decoder Models for Code Understanding and Generation
Projects
Neutral Network
GPT-4 powered YouTube comment bot
Tokenised Collateral Management
BSV blockchain automated settlement
XRPLoyalties
XRPL artist royalty distribution
BaysOpt for Bacterial ABS
Bayesian optimization for biological sims
Writings →
More coming soon...