Bachelor's degree in Computer Science, Computer Engineering, Electrical Engineering, or equivalent experience
5+ years of experience in designing scalable backend systems
Strong understanding of LLM inference mechanics
Experience with Kubernetes/Ray and containerization
Proficient in C++ and Python
Skilled in debugging, profiling, and system-level performance optimization
Effective communicator with mentoring skills
Responsibilities
Lead the technical direction for the model serving platform
Build core components like execution runtimes and distributed inference systems
Develop high-performance C++ and CUDA/HIP modules
Collaborate with ML researchers for productionizing multimodal models
Build Python APIs to expose model capabilities
Mentor engineers through code reviews and design discussions
Drive performance profiling and benchmarking across the stack
Benefits
Medical, dental, and vision insurance
401k plan
Daily lunch, snacks, and beverages
Flexible time off
Competitive salary and equity
Full Job Description
About the role
This is a rare chance to help architect and lead the development of Sciforium's next-generation model serving platform, the high-performance engine that will bring a multimodal, highly efficient foundation model to market. As a senior technical leader, you'll not only build core components yourself but also guide and mentor other engineers, influencing engineering direction, standards, and execution quality.
You will learn and shape the full AI stack: from GPU kernels and quantized execution paths to distributed serving, scheduling, and the APIs that power real-time AI applications. If you enjoy deep systems work, thrive on ownership, and want to lead engineers in building foundational AI infrastructure, this role puts you at the center of SciForium's mission and growth.
What you'll do
Lead the technical direction of the model serving platform, owning architecture decisions and guiding engineering execution.
Build core serving components including execution runtimes, batching, scheduling, and distributed inference systems.
Develop high-performance C++ and CUDA/HIP modules, including custom GPU kernels and memory-optimized runtimes.
Collaborate with ML researchers to productionize new multimodal models and ensure low-latency, scalable inference.
Build Python APIs and services that expose model capabilities to downstream applications.
Mentor and support other engineers through code reviews, design discussions, and hands-on technical guidance.
Drive performance profiling, benchmarking, and observability across the inference stack.
Ensure high reliability and maintainability through testing, monitoring, and engineering best practices.
Troubleshoot and resolve complex issues across GPU, runtime, and service layers.
Ideal candidate profile
Bachelor's degree in Computer Science, Computer Engineering, Electrical Engineering, or equivalent practical experience
5+ years of experience designing and building scalable, reliable backend systems or distributed infrastructure.
Strong understanding of LLM inference mechanics (prefill vs decode, batching, KV cache)
Experience with Kubernetes/Ray, Containerization
Strong proficiency in C++, Python.
Strong debugging, profiling, and performance optimization skills at the system level.
Ability to collaborate closely with ML researchers and translate model or runtime requirements into production-grade systems.
Effective communication skills and the ability to lead technical discussions, mentor engineers, and drive engineering quality.
Comfortable working from the office and contributing to a fast-moving, high-ownership team culture.
Nice-to-have
Experience with ML systems engineering, distributed GPU scheduling, open source inference engine like vLLM, Sglang, or TRT-LLM
Experience in building large scale ML/MLOps infrastructure
Proficiency in CUDA or ROCm and experience with GPU profiling tools
Experience at an AI/ML startup, research lab, or Big Tech infrastructure/ML team.
Familiarity with multimodal model architectures, raw-byte models, or efficient inference techniques.
Contributions to open-source ML or HPC infrastructure
