AI Accelerator Software Engineer

Location: Ramat Hahayal, Tel Aviv

Company: GSI Technology Israel

Employment Type: Full-time

About GSI Technology

GSI Technology is building Gemini2, an advanced Associative Processing Unit designed for efficient AI execution with ultra-low latency, high parallelism, and low power consumption.

The Role

We are looking for exceptional engineers who enjoy working close to the hardware and extracting maximum performance from novel compute architectures.

As an AI Accelerator Software Engineer, you will develop highly optimized low-level software, mathematical kernels, and execution flows for GSI’s Associative Processing Unit.

You will work on performance-critical code involving parallel execution, memory movement, DMA behavior, instruction scheduling, synchronization, and throughput optimization.

This role combines systems programming, accelerator kernel development, hardware-aware algorithm design, and AI inference optimization.

You will collaborate closely with architecture, compiler, and AI teams to shape how next-generation AI workloads execute on custom hardware.

What You’ll Work On

Design and optimize low-level compute kernels for AI and signal-processing workloads

Develop optimized implementations for: Transformer inference, LLM/VLM execution flows, OpenCV pipelines, FFTs, Edge AI workloads, Optimize memory access patterns, DMA utilization, execution flow, and scheduling

Analyze bottlenecks using profilers, traces, and hardware analyzers , Prototype and benchmark new execution strategies

Build internal tooling, testing infrastructure, and performance analysis frameworks

Use modern AI-assisted engineering tools to accelerate development and exploration

Required Qualifications

B.Sc. or M.Sc. in Computer Science, Electrical Engineering, Software Engineering, or related field

6+ years of professional C/C++ development in low-level, embedded, systems, firmware, accelerator, or performance-critical software

Strong background in:

Computer architecture

Memory hierarchies, caches, DMA, and bandwidth optimization

Parallel systems and performance-critical software

Hardware-aware algorithm optimization

Bit-level and systems-oriented reasoning

Preferred Experience

Accelerator programming (GPU, NPU, DSP, FPGA, or custom processors)

Assembly or other low-level programming

Development of compute kernels, firmware, or drivers

AI systems, deep learning infrastructure, or inference optimization

Profiling, tracing, benchmarking, and performance-debug tools

Effective use of modern AI development tools and coding assistants

Strong Fit Backgrounds

GPU kernels

DSP algorithms

Embedded high-performance C/C++

Computer architecture or RTL

AI inference optimization

Signal-processing or vision pipelines

Hardware/software co-design