Papers
Short research papers I have researched and written myself. Browse by category or search, and read the original PDF right here.
Performance Comparison of AI Agent, Data Analysis, and Machine Learning Workloads Across Python 3.9–3.15 Interpreter Versions
Jaeho Lim· NextGen AI Co.
We installed CPython 3.9–3.15 with the same build chain and measured 13 workloads. Standard-library workloads did not become faster after 3.11, and lazy imports in 3.15 reduced CLI startup time by a factor of 10.69.
Serverless Instance Freezing and Postgres Pooler Client Slot Exhaustion: A Next.js and Supabase Production Case and Local Reproduction
Jaeho Lim· NextGen AI Co.
We reconstructed the connection failures of an internal Next.js, Vercel, and Supabase tool from 11 incident records and 16 commits. A reproduction experiment confirmed that slot occupancy on frozen instances follows W·min(q,m), and showed that a release configuration prevents exhaustion.
Comparing AI Agent Runtimes on On-Premise Linux: Measuring Startup, Memory, CPU, and Latency of Node.js, Bun, Deno, and a Rust Native Binary
Jaeho Lim· NextGen AI Co.
We implemented the same agent loop in JavaScript (Node.js, Bun, Deno) and Rust and measured both. At 500 concurrent sessions, Rust's peak RSS was 159.7 MB, smaller than Node.js's 775.8 MB, and its median orchestration overhead was 2.24 ms.