4×
faster inferenceLLM threat detection on the AI Security team.
- Worked with the AI Security team to migrate a prompt-scan engine to state-of-the-art detection models for LLM threat analysis, including implementation, testing, debugging, and rollout support.
- Implemented chunking behavior for an inference engine in Rust and Go, improving support for long prompts and document-style inputs in AI security workflows.
- Introduced dynamic chunking to the inference engine, speeding up inference by 4× and cutting energy usage by 62% while raising throughput.
- Owned design, implementation, testing, and AWS deployment of a backend service that retrieves company profiles and market data through MCP-based integrations for cybersecurity research workflows.
- Rust
- Go
- AWS
- MCP
- LLM security
