LLM · RAG
Retriever-X
Hybrid retrieval pipeline with reranking for enterprise search.
PyTorchFAISSFastAPI
+18% recall · 40ms p95
Machine Learning Developer
I build production ML systems — from research prototypes to low-latency inference. Currently shipping LLM tooling and MLOps platforms at a small, fast team.
12
Models shipped
4.2B
Tokens served
9ms
p99 latency
6
Publications
LLM · RAG
Hybrid retrieval pipeline with reranking for enterprise search.
+18% recall · 40ms p95
Vision · Edge
Quantized detection model running on-device for quality control.
0.91 mAP · 3× faster
MLOps · Infra
Reproducible training orchestration with automatic model registries.
-60% deploy time
Languages & frameworks
LLMs & MLOps
Senior ML Engineer · Northwind AI
Leading LLM inference and retrieval systems.
ML Engineer · Helios Labs
Built vision pipelines and MLOps tooling.
Research Engineer · Cartograph
Geospatial models and data infrastructure.
© 2025 Derryl Kevin · Built with care