Software Engineer

Location: Ramat Hahayal, Tel Aviv

Employment Type: Full-time

Hybrid .

Who we are :

GSI Technology – A publicly traded, international high-tech company (NASDAQ: GSIT)

GSI is pioneering the Gemini APU—a cutting-edge, game-changing processor designed to accelerate compute-intensive tasks like large language models, machine learning, advanced image processing, and radar imaging.

If you're passionate about architecting high-performance software systems, implementing advanced algorithms, and drilling into low-level technical details, this is the role for you.

We’re seeking a dynamic and fast-learning engineer with a passion for diving deep into large language model implementations, and a keen focus on performance optimization and efficient execution.

We are looking for a SW engineer to implement AI models (CNN, LLM, VLM) with C++. This roll sits at the intersection of AI Algorithm and our cutting edge APU Architecture .

At GSI Technology, we don't just use AI chips—we build them. We are developing the Gemini® APU, a compute-in-memory processor designed to accelerate massive workloads like LLMs and VLMs.

We are looking for an architecturally-minded engineer who looking to combine understanding of LLM models and hardware constraints and can take complex AI algorithms (written in Python) and design them to run at lightning speed on our proprietary AI chip (using C/C++).

Who you are :

• You are a Software Engineer who is curious about state of the art AI models – how they work and how to implement them
• You want to deep dive and understand what happens "under the hood" when software talks to hardware (memory management, caching, latency).
• You are "Open Minded" , enjoy tackling unsolved problems and aren't afraid to learn a new proprietary architecture from scratch.

Key Responsibilities:

• Bridge the Gap: Translate high-level AI models (LLMs, Vision Transformers) from Python reference code into C/C++ implementations for the Gemini APU.
• System SW Architecture: Design and maintain the software stack and libraries that allow customers to run AI workloads on our AI chip.
• Performance Tuning: Squeeze every bit of performance out of the chip. Deal with memory bandwidth, latency, and compute constraints.
• Algorithm Engineering: Adapt standard algorithms (CNNs, Transformers) to fit a unique compute-in-memory architecture.
• Developing software compilers AI tools.
• Owning our SW device SDK

Must-Have Qualifications:

• M.Sc. / B.Sc. in Computer Science or Electrical Engineering.
• 7+ years of active full time hands-on software development (of complex and innovative SW architecture)
• Algorithmic Mindset: You can look at a mathematical formula or an algorithm and understand how to implement it efficiently in code.
• SW Architecture Knowledge

Willing to learn and grow into: (We will teach you)

• Focus on C/C++ and Python.
• Understanding of how processors work
• Understand registers, memory hierarchy, and pointers.
• Understanding of how processors work (SIMD, Vector processing, Caches, Pipelines).
• Understanding of Deep Learning primitives (Quantization, Pruning, Matrix Multiplication).
• Experience writing compilers, assemblers, or SDKs.

Advantages:

• Graph compilers
• Graph optimizations
• Tiling/slicing strategies
• Scheduler tuning
• Understanding of LLM architecture
• Experience with CUDA, OpenCL, or Writing Kernels.
• Background in Computer Vision or Signal Processing pipelines.

Privacy Statement: All applications will be handled with strict confidentiality. Your information will not be shared without your consent.