Walmart is building the next generation of AI powered experiences on smart TVs, including on-device large language models, real-time content understanding, privacy-preserving audience insights, and interactive generative graphics. The AdTech M&R data team needs to deliver reporting and measurement for Advertisers to analyze and optimize campaigns, solving sophisticated and high-impact problems at scale.
Requirements
- Proven track record in generative models (LLMs, diffusion, transformers, VAEs) and modern CV/NLP stacks.
- C++ / Rust and Python expertise; comfortable refactoring kernels for efficiency on limited memory devices.
- Hands on experience with edge inference (TensorRT, CoreML, TVM, ONNX, TFLite, WebGPU, or similar).
- Familiarity with CTV / ACR pipelines (frame grabbers, fingerprinting, video embeddings) or embedded multimedia systems.
- Strong grasp of distributed training (DDP/Horovod) and MLOps (Kubeflow, Airflow, MLFlow, Feature stores).
- Knowledge of data privacy frameworks (FL, DPSGD, HE, SMPC) and ability to translate compliance constraints into code.
- Experience optimizing models for ARM Cortex‚ NEON, NPU, or VVC ASICs inside smart‚ TV SoCs.
Responsibilities
- Design & train GenAI models targeting CTV use cases: on device LLM quantization, multimodal video text encoders, and diffusion based visual experience generators.
- Deploy to edge environments optimize PyTorch/TensorFlow models with ONNX RT/TVM, Arm NN, or Qualcomm SNPE; own CI/CD pipelines that push updates to millions of TVs.
- Integrate privacy enhancing tech such as federated learning, encrypted feature extraction, and PSI to align with global data regulations and Walmart stack.
- Prototype novel user features (e.g., conversational search, generative ad creatives, adaptive picture modes) in collaboration with product, design, and ad tech teams.
- Publish & share file patents, present at CVPR/NeurIPS, mentor junior ML engineers, and contribute to open source edge AI tooling.
Other
- 5 years building and shipping ML or DL products (or 2 with a PhD degree).
- Bachelor or higher in CS, EE, Math, or related field.
- Background in ad tech or retail media optimization (incrementally real time ROAS auction dynamics).
- Familiarity with incrementality measurement, attribution, and real time auction dynamics (e.g., bid shading, floor price optimization).
- Working knowledge of data privacy frameworks (FL, DP SGD, HE, SMPC) and ID resolutions (UID2, RampID, SharedID).
