Concurrency, throughput & the latency tradeoff
When does increasing concurrency stop helping throughput and start hurting latency?
Projects & experiments
Start with a question.
Follow the evidence.
A collection of systems experiments and, as the work develops, open-source tools. Each project starts with a concrete question and makes its scope and limitations explicit.
AWS helpers
Tools for tracking EC2 resource changes and checking AWS volume status.
Easy Patcher
A tool for scanning and patching EC2 instances with Amazon SSM.
Security Watchdog
A tool that associates vulnerable Amazon packages with affected infrastructure instances in a report.
Gate Keeper
A Chrome extension for dynamic malware analysis.
ProtectMe
A Python privacy-advisor application focused on publicly available Facebook data and suggested privacy settings.
Hello Bud
A voice and text assistant for personalized academic and career planning.
Questions on my list
Where I want to dig deeper
These are directions for exploration, not completed projects.
- Model serving
How does concurrency affect throughput and tail latency?
- RAG
How do stale documents or retrieval failures affect answers? Can we detect the failures?
- Context engineering
When does extra context help, and when does it increase cost or reduce task success?
- Agent loops
How do we bound retries, runtime, and cost? What happens after a tool failure?
- Graph-based workflows
Can execution resume safely after a crash without repeating side effects?