Founding Engineer -- AI Inference Stack
About the role
Founding Engineer role focused on building the AI inference stack and low-latency software/hardware integration at Piris Labs. Internship-to-founding equity path with potential Series A alignment.
Responsibilities
- Architect the Stack: Iterate on our software layer that orchestrates inference across heterogenous cluster of compute resources.
- Kernel and Compiler Optimization: Write and optimize high-performance kernels in CUDA, Triton, or custom targets to squeeze every drop of performance from the system.
- The Runtime: Build the low-latency inference server, think a more performant custom version of vLLM or TensorRT-LLM, that manages KV cache at scale without the overhead of traditional PCIe bottlenecks.
- Voice and Agentic Optimization: Solve the unique challenges of instant-on Voice AI, focused on latency, and the high-context demands of coding agents, focused on memory management.
Qualifications
- Curiosity-driven, with a genuine passion for compute architectures and problem solving
- Systems Obsessed: deep understanding of computer architecture, memory hierarchies, and low-level systems programming in C++, Rust, or CUDA.
- AI Fluent: understanding transformer architectures and experience with inference frameworks like vLLM, TensorRT, ONNX, and Kubernetes.
- A First-Principles Thinker: not afraid to challenge standard approaches for 10x gains.
Skills
Benefits
- The Path to Full Time: This internship is a trial run for a founding-level equity stake. We are looking for people to grow with us through our Series A and beyond.
About Piris Labs
Cerebras-speed inference but scalable
Piris Labs offers a full-stack inference service that eliminates the AI data movement bottleneck. By pairing proprietary photonic hardware with a vertically optimized software stack, we minimize the "memory wall" associated with expensive GPUs. This allows us to deliver the same performance as traditional clusters at a fraction of the cost. Our technology improves effective FLOP utilization and reduces latency, finally making the unit economics of trillion-parameter models sustainable. Founded by a team of MIT physicists and Meta AI experts.