Miller Juma
System Performance Engineer at Microsoft
About
I'm a System Performance Engineer at Microsoft, specializing in benchmarking, profiling, and optimizing large-scale systems that power experiences for millions of users worldwide. My work centers on extracting maximum performance from infrastructure—identifying bottlenecks, measuring what matters, and making systems faster, more efficient, and more reliable.
With deep expertise in performance analysis and benchmarking methodologies, I've optimized critical systems that process billions of requests daily, reducing latency, cutting costs, and improving resource utilization. My approach combines rigorous measurement with data-driven insights to deliver tangible improvements. Beyond performance engineering, I maintain a strong curiosity in machine learning—exploring how ML can inform and enhance performance optimization strategies.
Experience
System Performance Engineer
PresentMicrosoft
- Lead performance benchmarking and optimization for large-scale distributed systems processing 10M+ transactions daily
- Developed comprehensive benchmarking frameworks and methodologies to measure system performance across multiple dimensions
- Drove performance initiatives that reduced P95 latency by 65%, improved throughput by 80%, and cut infrastructure costs by $2M+ annually
- Implemented advanced profiling and monitoring solutions providing real-time insights into system bottlenecks and resource utilization
- Collaborated with engineering teams to establish performance SLAs and optimize critical paths in high-traffic services
- Pioneered data-driven performance analysis using statistical methods and ML-based anomaly detection for capacity planning
Selected Projects
Enterprise Benchmarking & Performance Testing Framework
Designed and implemented a comprehensive benchmarking platform for evaluating system performance across distributed services. The framework automates load generation, resource monitoring, and statistical analysis, enabling teams to measure throughput, latency, and resource utilization under various workloads. Benchmarking results drove optimization efforts that improved overall system performance by 45% and identified critical bottlenecks before production deployment.
High-Performance Distributed Caching Optimization
Profiled and optimized a globally-distributed caching layer serving 150K+ RPS, reducing P95 latency from 45ms to 8ms. Through systematic benchmarking, identified hot keys, analyzed memory access patterns, and optimized serialization protocols. Implemented advanced monitoring and alerting that reduced cache-related incidents by 85%. The optimizations saved $1.5M annually in infrastructure costs while improving user experience across all regions.
ML-Driven Performance Anomaly Detection
Built an ML-powered system to detect performance anomalies and predict capacity issues before they impact users. Applied statistical analysis and machine learning models to historical performance metrics, identifying subtle degradation patterns that traditional monitoring missed. The system provides early warnings for performance regressions, enabling proactive optimization and reducing P95 incident response time by 60%. This project combines my performance engineering expertise with curiosity-driven exploration of ML applications.
Technical Skills
Performance & Benchmarking
Load Testing, Profiling, Performance Analysis, Capacity Planning, Latency Optimization, Throughput Tuning
Monitoring & Observability
Prometheus, Grafana, Azure Monitor, Application Insights, Distributed Tracing, Metrics Analysis
Languages & Tools
C#, Python, SQL, PowerShell, Bash, Performance Testing Frameworks, Statistical Analysis
Systems & ML (Curiosity-Driven)
Distributed Systems, Azure, Kubernetes, Machine Learning, Anomaly Detection, Data Analysis
Get in Touch
I'm always open to discussing performance optimization challenges, benchmarking methodologies, and opportunities to make systems faster and more efficient. Whether it's tackling complex performance bottlenecks, designing benchmarking frameworks, or exploring ML applications in performance engineering—let's connect.