AI Software Engineer

About the role

ABOUT ELASTIX AI

We are building the next-gen AI inference platform.

DESCRIPTION

Job Title: Software Engineer, AI Inference Platform

Company: ElastixAI, Inc.

Location: Seattle, WA (Hybrid - 3 days/week in office)

ABOUT ELASTIXAI

ElastixAI is an early-stage startup building the next-generation AI inference infrastructure — co-designed across ML software and custom accelerator hardware. Our platform dynamically optimizes inference efficiency and scalability across diverse deployments, enabling adaptive, high-performance AI serving.

ROLE SUMMARY

We’re looking for a systems-minded AI Software Engineer to join our core inference platform team. You’ll design and extend the low-level serving stack — hacking open-source frameworks like vLLM, SGLang, and TensorRT-LLM, building new model sharding and scheduling logic, and integrating deeply with our proprietary AI accelerator. This role sits at the intersection of ML systems, compiler/runtime engineering, and hardware-software co-design.

KEY RESPONSIBILITIES

- Architect, extend, and optimize core components of our AI serving platform for throughput, latency, and scalability.

- Customize open-source serving frameworks (e.g., vLLM) for proprietary model ingestion and accelerator integration.

- Develop efficient model partitioning, scheduling, and memory management strategies for multi-device inference.

- Collaborate with ML engineers on model export and runtime optimization (quantization, graph transforms).

- Work closely with hardware engineers to influence accelerator interface design and performance tuning.

- Build APIs and runtime tools enabling flexible, PyTorch-native model deployment on our infrastructure.

- Profile, debug, and optimize across the full stack — from Python orchestration to C++ kernels and PCIe drivers.

REQUIRED QUALIFICATIONS

- BS/MS/PhD in Computer Science, Electrical/Computer Engineering, or related field.

- 3+ years of professional experience in systems programming, ML infrastructure, or distributed inference.

- Proficient in C++ and Python, with strong debugging and performance analysis skills.

- Deep familiarity with one or more LLM serving frameworks (vLLM, SGLang, TensorRT-LLM, DeepSpeed-Inference, etc.).

- Understanding of model deployment internals — token scheduling, KV caching, batching, and pipelined inference.

- Comfortable working close to the hardware abstraction layer — CUDA, PCIe, memory management, or runtime scheduling.

- Strong collaboration and communication skills; ability to work cross-functionally in a fast-paced startup environment.

PREFERRED / BONUS

- Experience with hardware-aware ML optimization, compiler/runtime integration, or accelerator SDKs.

- Hands-on experience profiling GPU/accelerator workloads.