Joseph Mburu

Staff-level AI Systems Engineer with 6+ years shipping production ML infrastructure at scale. Specialized in sub-100ms distributed inference (10M+ daily requests, 99.99% SLA) and federated learning platforms serving Fortune 500 clients.

Rare hybrid: production-grade Rust systems + state-of-the-art NLP. 3 patents, 2.5k+ GitHub stars, speaker at NeurIPS/MLSys workshops.

Staff AI Engineer - QuantumDynamics AI - Remote

(2021-01)

Architected and shipped mission-critical ML systems for healthcare and enterprise clients:

• Federated Learning Platform (NVIDIA) — Built Rust-based networking layer reducing cross-region latency 58% (P99: 340ms → 143ms). Enabled HIPAA-compliant model training across 12 global nodes without data centralization.
• Real-Time Inference API — Scaled from 100K to 10M+ daily predictions via custom Kubernetes operators and GPU autoscaling. Maintained 99.99% uptime during 3x traffic spikes.
• Florida Medicaid Fraud Detection — Led 3-engineer team shipping gradient-boosted ensemble with real-time feature store. 98.2% precision at 94% recall; $12M annual fraudulent claims flagged for review.
• NLP Latency Optimization — Reduced production transformer latency 40% via ONNX Runtime + custom quantization kernels. Serves 1M+ daily active users.
• Engineering Culture — Mentored 4 engineers (2 promoted to Senior); established code review standards and incident response playbooks adopted company-wide.

Machine Learning Engineer (Contract) - JAXA Labs, Jacksonville - Jacksonville, FL

(2020-01 - 2021-01)

Built production computer vision systems for maritime logistics:

• Deployed real-time object detection at PortMiami (YOLOv5 → TensorRT), processing 30 FPS across 12 camera streams for automated cargo tracking.
• Co-authored 2 patents on adverse-weather object detection (USPTO pending: maritime fog/rain robustness).
• Reduced false-positive rate 67% via synthetic data augmentation pipeline.

Research Engineer - University of Florida AI Lab - Gainesville, FL

(2018-01 - 2019-01)

Shipped open-source tools bridging research and production:

• PolyglotSentiment — Multilingual sentiment analysis library (Python/Rust). 2,500+ GitHub stars; adopted by 3 startups for production social media monitoring.
• Thesis work on few-shot learning for low-resource African dialects (Wolof, Swahili); 4x improvement over mBERT baseline with <100 labeled examples.

MS - Data Science - California State University, Los Angeles (2020-01 - 2022-01)

BS - Computer Science - California State University, Stanislaus (2016-01 - 2020-01)