Open to opportunities

Kamran Razavi

@kamranrazavi

Message

I build AI datacenter simulators that optimize GPU capacity, scheduling, performance, and cost.

Germany

Message

What I'm looking for

I’m looking to build AI and cloud infrastructure that helps teams make better capacity-planning, scheduling, performance, and cost decisions through rigorous modeling and production-grade engineering.

I lead the design and development of NeonSim at turbalance GmbH, helping cloud providers size GPU/AI datacenters, compare scheduling strategies, and forecast performance, utilization, and cost before purchasing hardware.

I build GPU and Multi-Instance GPU resource models, AI inference-serving simulations, SLO-aware scheduling, and autoscaling capabilities. I’m also architecting GoldSim, a high-performance C++20 simulator for Linux x86_64 and ARM64 platforms.

My work combines mathematical modeling with production engineering across cloud platforms, containers, distributed systems, and machine learning frameworks. I build observability and validation tooling with Perfetto, HDF5, and benchmarking against published measurements.

I’ve also worked at Bloomberg, taught distributed systems and algorithms, supervised student research, reviewed pull requests, and partnered directly with customers such as StackIt to turn infrastructure-planning needs into simulation capabilities.

Experience

Work history, roles, and key accomplishments

Current

R&D System Engineer

Current

Turbalance GmbH

Jan 2025 - Present (1 year 7 months)

Lead the design and development of NeonSim, a Java discrete-event simulator for GPU/AI datacenters, and architect GoldSim, a high-performance C++20 unified simulator. Partner with customers to translate infrastructure-planning requirements into simulator capabilities.

Java C++Python Kubernetes Docker AWS PyTorch TensorFlow Gurobi CI CD Protobuf Perfetto CMake LLVM Clang

Software Engineer Intern

Bloomberg L.P.

Jan 2018 - Dec 2018 (11 months)

Created a Docker-based simulator for A/B testing of package version control in production environments. Enhanced deployment efficiency and reliability by optimizing version testing workflows.

Docker A B Testing Version Control Deployment Reliability

Education

Degrees, certifications, and relevant coursework

Technical University of Darmstadt

Doctor rerum naturalium, Computer Science

2019 - 2024

Activities and societies: Lecturer for two master courses; supervised master's and bachelor's theses.

Pursued a Doctor rerum naturalium in Computer Science, focusing on resource-efficient inference serving with SLO guarantees. Authored publications in RTAS, EuroMLSys, JSys, and EuroP4, and received the ISCC'24 Paper Award.

Maynooth University

Master of Science, Dependable Software Systems

2017 - 2018

Activities and societies: Erasmus Mundus Partner Country Scholarship recipient.

Completed an Erasmus Mundus Double Master of Science in Dependable Software Systems, with a thesis on mining historical data using machine learning and graph algorithms. Awarded the Erasmus Mundus Partner Country Scholarship.

University of St Andrews

Master of Science, Dependable Software Systems

2016 - 2017

Activities and societies: First place in Bloomberg CodeCon; fifth in ICPC Scotland.

Completed part of the Erasmus Mundus Double Master's program, with a thesis on Kubernetes blue-green deployment. Won first place in Bloomberg CodeCon and fifth in ICPC Scotland.

University of Tehran

Bachelor of Science, Mathematics and its Applications

2011 - 2016

Activities and societies: Top one percent in the Iranian national university entrance exam.

Earned a Bachelor of Science in Mathematics and its Applications, with a thesis on optimizing cloud storage using MRC. Ranked in the top one percent of the Iranian national university entrance exam.