Engineer, Machine Learning Tools

Qualcomm seeks an Engineer, Machine Learning Tools in Hyderabad to develop and commercialize the Qualcomm AI Runtime (QAIRT) SDK for Qualcomm SoCs. The Generative AI team integrates models on Qualcomm chipsets and enables power-efficient, on-device inference without a cloud connection, including large language and vision models on Snapdragon platforms.

Responsibilities

• Improve inference performance for large models and deploy large C/C++ software stacks using software engineering best practices.
• Develop, modify, and validate embedded and cloud-edge software and collaborate with systems, hardware, architecture, and test engineers on system-level solutions and performance requirements.
• Apply knowledge of LLMs, Transformers, and edge-based Generative AI deployment to the AI runtime.

Requirements

• The role-specific requirements call for a bachelor’s or master’s degree in computer science or equivalent and 6+ years of relevant software-development experience. The posting also lists general minimum qualifications: a related bachelor’s degree with 3+ years of software engineering or related experience, a related master’s with 2+ years, or a related PhD with 1+ year; plus 2+ years of academic or work experience with a programming language such as C, C++, Java, or Python.
• Strong understanding of generative AI models, including LLMs, LVMs, and LMMs, and concepts such as self-attention, cross-attention, and KV caching. Knowledge of floating-point and fixed-point representations and quantization; experience optimizing algorithms for CPU, GPU, or NPU accelerators; strong C/C++, design-pattern, and operating-system knowledge; and Python scripting skills.
• Analytical, debugging, communication, and cross-team collaboration skills.

Preferred qualifications

• SIMD processor architecture, system design, object-oriented development, and SIMD kernel development; familiarity with Linux and Windows.
• Familiarity with llama.cpp, MLX, or MLC; knowledge of PyTorch, TFLite, or ONNX Runtime; and parallel-computing experience with OpenCL or CUDA.