Software Engineer, Applied Ai, Pixel

Google Google · Big Tech · New Taipei, Banqiao District, New Taipei City, Taiwan

Software Engineer role focused on selecting, evaluating, and fine-tuning state-of-the-art AI models for user-facing features in Pixel software. Responsibilities include owning the model-to-feature lifecycle, designing evaluation pipelines, and collaborating with research and product teams. The role involves deploying and optimizing LLMs and multi-modal models for on-device and cloud execution, staying current with Generative AI advancements, and translating research into product features.

What you'd actually do

  1. Contribute to the strategy for selecting, fine-tuning, and deploying foundational models to achieve specific product goals.
  2. Work on the deployment and optimization of large language models (LLMs) and multi-modal models for both on-device and cloud execution. This includes leveraging and improving ML infrastructure for model evaluation, data processing, and debugging.
  3. Contribute to the design and implementation of data collection and processing solutions. Develop evaluation frameworks to measure the real-world performance and impact of machine learning models on the user experience.
  4. Stay current with the latest advancements in Generative AI. Translate cutting-edge research into tangible product features that create transformative and helpful experiences for Pixel users.

Skills

Required

  • software development in Python, Java or C++
  • GenAI techniques (e.g., LLMs, multi-modal, large vision models)
  • GenAI-related concepts (e.g., language modeling, computer vision)

Nice to have

  • Master's degree in Computer Science, specializing in Machine Learning, Artificial Intelligence, or Natural Language Processing
  • independently building and contributing to the design of user-facing GenAI features
  • JAX, Tensorflow, PyTorch
  • ML fundamentals
  • data engineering skills
  • modern ML model evaluation techniques
  • industry-standard benchmarks and metrics

What the JD emphasized

  • state-of-the-art AI models
  • user-facing features
  • Pixel software
  • model-to-feature lifecycle
  • robust evaluation pipelines
  • on-device deployment
  • large language models (LLMs)
  • multi-modal models
  • Generative AI

Other signals

  • fine-tuning state-of-the-art AI models
  • develop innovative, user-facing features
  • design and implement robust evaluation pipelines
  • on-device deployment
  • deploy and optimization of large language models (LLMs) and multi-modal models