Website:
Job details:
Member of Technical Staff | AI Systems, Kernels & Compilers
Location: India | MANDATORY ONSITE
Employment: Full time
Compensation: US$90,000 per year
Relocation bonus: Offered
Experience: No prior professional experience required
Supayou is building a frontier AI lab in India, currently operating in stealth. We began as a software engineering services company, building products and solving complex problems for partners. Today, our mission is to make powerful intelligence accessible at an affordable cost, on the devices people already use. We’re working toward a future where advanced AI can run efficiently with limited memory, compute, power, and connectivity, with ambitions that extend to hardware as small as a Raspberry Pi.
As a Member of Technical Staff, you’ll work closely with engineers and researchers on the systems that make efficient AI possible. The work spans model inference, GPU kernels, compiler development, and deployment across different hardware platforms. You’ll help turn research ideas into working implementations, measure their performance, investigate bottlenecks, and build reliable software around them.
Depending on your strengths and interests, you’ll develop and optimize GPU kernels in Triton, including FlashAttention style attention kernels, matrix multiplication, and fused operations. You’ll contribute to compiler passes, lowering pipelines, and code generation using LLVM and MLIR to improve how models execute on target hardware.
You’ll also work on inference latency, throughput, memory usage, and data movement across CPU and GPU execution paths. This includes exploring quantization, efficient KV cache management, and techniques for running models within limited compute and memory budgets. Profiling tools, benchmarks, correctness tests, inference runtimes, and deployment infrastructure will be part of the engineering work.
We’re looking for strong software engineering fundamentals: clear reasoning, sound knowledge of data structures and algorithms, clean code, debugging skills, and an understanding of how software interacts with hardware. You should be comfortable building in a language such as C++, Python, or Rust, and willing to deepen your understanding of memory management, concurrency, operating systems, and computer architecture.
Experience with Triton, CUDA, FlashAttention, LLVM, MLIR, PyTorch, embedded systems, or inference optimization is welcome. Your strengths can be in one area, with curiosity and a willingness to learn across the others. We value people who can explain their decisions, test their assumptions, and follow a difficult problem through to a working solution.
We have open headcount, and prior professional experience is not required. Students, recent graduates, self taught engineers, researchers, and experienced builders are welcome. If you’re building something you’re proud of and want to contribute to this mission, we’d love to see your work, whether that’s a personal project, a research experiment, an open source contribution, or a system you built from scratch.
This role requires working onsite with the team in India. We offer a relocation bonus to support your move.
Our hiring process consists of one take home assignment followed by one interview loop with a panel from the team. There are no LeetCode rounds. We’ll discuss your work, your technical decisions, and how you approach engineering problems.
- Come help shape Supayou’s research, engineering, and culture as we build from India for the world.
Click on Apply to know more.