Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

I'm an AI engineer with a decade of experience deploying LLM inference and production ML platforms. I design and operate reliable, scalable AI services using Kubernetes, Helm, ArgoCD, and modern cloud infrastructure to minimize latency and maximize uptime across multi-cluster environments. I excel at Python automation, observability, and CI/CD, delivering fast, measurable SLA improvements. I thrive in fast-paced teams and enjoy leading initiatives from inception to production, continuously optimizing performance and cost.…I'm an AI engineer with a decade of experience deploying LLM inference and production ML platforms. I design and operate reliable, scalable AI services using Kubernetes, Helm, ArgoCD, and modern cloud infrastructure to minimize latency and maximize uptime across multi-cluster environments. I excel at Python automation, observability, and CI/CD, delivering fast, measurable SLA improvements. I thrive in fast-paced teams and enjoy leading initiatives from inception to production, continuously optimizing performance and cost.

Andrei Jaume-Willenska

AI Engineer, Data Scientist, DevOps Developer, +2





I'm an AI engineer with a decade of experience deploying LLM inference and production ML platforms. I design and operate reliable, scalable AI services using Kubernetes, Helm, ArgoCD, and modern cloud infrastructure to minimize latency and maximize uptime across multi-cluster environments. I excel at Python automation, observability, and CI/CD, delivering fast, measurable SLA improvements. I thrive in fast-paced teams and enjoy leading initiatives from inception to production, continuously optimizing performance and cost.…I'm an AI engineer with a decade of experience deploying LLM inference and production ML platforms. I design and operate reliable, scalable AI services using Kubernetes, Helm, ArgoCD, and modern cloud infrastructure to minimize latency and maximize uptime across multi-cluster environments. I excel at Python automation, observability, and CI/CD, delivering fast, measurable SLA improvements. I thrive in fast-paced teams and enjoy leading initiatives from inception to production, continuously optimizing performance and cost.

Available to hire

I’m an AI engineer with a decade of experience deploying LLM inference and production ML platforms. I design and operate reliable, scalable AI services using Kubernetes, Helm, ArgoCD, and modern cloud infrastructure to minimize latency and maximize uptime across multi-cluster environments.
I excel at Python automation, observability, and CI/CD, delivering fast, measurable SLA improvements. I thrive in fast-paced teams and enjoy leading initiatives from inception to production, continuously optimizing performance and cost.

Skills

Experience Level

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Intermediate

Intermediate

Intermediate

Intermediate

Intermediate

Language

English

Fluent

Javanese

Advanced

Bashkir

Intermediate

Aragonese

Advanced

Afar

Intermediate

Work Experience

AI Engineer at Virtu

October 1, 2024 - February 28, 2026

Integrated LLM inference APIs into production platforms, reducing request latency by 28% and increasing concurrent inference capacity to support 4x client traffic. Managed Kubernetes clusters and Helm charts with autoscaling and resource quotas across staging and production, improving cluster utilization by 33% and reducing costs by 18%. Built ArgoCD GitOps pipelines and CI/CD integrations to enable rapid model rollouts. Implemented observability with Prometheus, Grafana, and ELK logging, decreasing MTTR by 55%. Automated Python scripts for inference orchestration and API health checks, cutting manual checks by 90% and bolstering SLA compliance. Troubleshot production LLM scaling issues and implemented adaptive batching to restore service capacity during high-demand incidents.

Senior Infrastructure Engineer at SparkCognition

June 1, 2017 - October 1, 2024

Led design and deployment of Kubernetes inference clusters, reducing model-serving latency by 35% and increasing throughput 3x in global production. Implemented Helm charts and ArgoCD pipelines for reproducible LLM deployments, accelerating release cycles by 60% with zero-downtime rollbacks. Automated Python-based deployment tooling and inference API integrations, reducing manual interventions by 80% and cutting MTTR by 40%. Refactored cloud IaC to reduce operational costs by 22% and improve reliability. Built observability stack with Prometheus, Grafana, and ELK, introducing SLO-based alerts and lowering incident detection time by 70% across services. Optimized LLM inference with batching and quantization, cutting GPU costs by 45% while maintaining 93% model accuracy.

Robotics Developer at Austin Club

August 1, 2014 - May 1, 2015

Developed ROS-based navigation modules in C++ and Python, improving autonomous path planning accuracy and reducing localization drift by 40% in indoor trials. Designed sensor fusion pipelines with LiDAR and IMU, increasing obstacle detection range by 25% and enhancing real-time decision making. Implemented automated build and CI for embedded firmware using Make and Jenkins, reducing integration errors and test time by 35%. Led a team to prototype vision-based object recognition achieving 87% classification accuracy under varied lighting. Optimized motor control for energy efficiency, extending run time by 18%. Documented system architecture and mentored new members to accelerate onboarding.

Robotics Engineer at E1K Robotics

August 1, 2014 - May 1, 2015

Developed ROS-based navigation modules in C++ and Python, improving autonomous path planning accuracy and reducing localization drift by 40% in indoor trials. Designed sensor fusion pipelines integrating LiDAR and IMU data, increasing obstacle detection range 25% and enhancing real-time decision making for competitions. Implemented automated build and CI processes for embedded firmware using Make and Jenkins, reducing integration errors and test time by 35%. Led a team of four students to prototype vision-based object recognition, achieving 87% classification accuracy under varied lighting conditions in competition scenarios. Optimized motor control algorithms to improve energy efficiency, extending run time by 18% during continuous autonomous operation in test deployments.