Senior AI Application Engineer – LLM, VLM Enablement

This is a fully remote position, open to applicants in Netherlands.

📋 Description

• Take charge of LLM and VLM support on Metis and Europa AIPUs, encompassing the exploration of new models through to their deployment in the Voyager SDK.

• Investigate new families and architectures of LLM and VLM, prototype and evaluate them on Metis and Europa, and ascertain their operational requirements on the hardware.

• Convert research findings into Voyager SDK features that are supported with stable APIs, thorough testing, documentation, and release protocols.

• Manage PyTorch references involving quantization, weight mappings, and contracts that govern inputs, outputs, weights, and persistent states.

• Provide model-side contracts to the compiler team, ensuring that runtime and tests adhere to the same definitions.

• Incorporate LLM and VLM serving into the Voyager SDK pipeline builder, including prefill, decoding, KV-cache management, streaming text output, and multimodal inputs.

• Assess accuracy against floating-point references and either address or clarify discrepancies.

• Enhance the model enablement platform for LLM and VLM deployment, which includes compiler, runtime, and hardware environments, as well as closed-loop runs and evaluations.

• Suggest Voyager SDK capabilities for GenAI, agents, and speech applications on edge devices, while identifying representative model-zoo demonstrations.

• Utilize AI coding agents to construct, test, and validate work.

⛳️ Requirements

• A minimum of 5 years of software engineering experience.

• At least 2 years focused on ML inference, ML systems, or ML tooling.

• Strong comprehension of transformer inference, including attention variations, KV caching, prefill versus decode, sampling, tokenization, and VLM vision-language integration.

• Practical knowledge of LLM quantization, covering weight-only and weight-activation schemes, calibration, and outlier management.

• Capability to differentiate quantization loss from implementation errors.

• Proficient in Python and PyTorch.

• Experience in delivering production-quality software.

• Regular usage of AI coding agents on actual codebases, with the discernment to validate or dismiss their outputs.

• Ability to translate research code and hardware limitations into viable designs.

• Preferred: experience with inference engines or serving stacks such as llama.cpp, vLLM, ONNX Runtime GenAI, or ExecuTorch.

• Preferred: background in kernel development, tensor compilers like Triton, TVM, or MLIR, or profiling models on accelerator hardware.

• Preferred: expertise in speech applications (ASR, TTS) or agent-based applications on embedded devices.

• Preferred: experience with evaluation or testing frameworks for ML systems or AI agents, or tool-calling agents like MCP servers.

🏝️ Benefits

• Competitive compensation package.

• Pension plan.

• Comprehensive employee insurance coverage.

• Opportunity to acquire company shares.

• Flexible work arrangements.

• An open culture that fosters creativity and continuous innovation.

• Collaborative ownership and autonomy accompanied by accountability.

• An equal-opportunity, welcoming, and inclusive workplace.