Machine Learning Engineer 4

Adobe Adobe · Enterprise · Noida, India

Machine Learning Engineer at Adobe Genuine Engineering team, focused on building and deploying deep learning models for fraud detection and account protection. The role involves managing the full model lifecycle, from data engineering and architecture development to large-scale GPU training, deployment, and monitoring. The team is developing in-house behavioral foundation models for identity-preserving representations.

What you'd actually do

  1. Build and train deep learning models from scratch, including custom transformer and attention-based architectures for long behavioral event sequences.
  2. Own the full training stack: event tokenization, temporal and positional embeddings, self-supervised pretraining (e.g., masked modeling, contrastive learning), and downstream fine-tuning.
  3. Train large models efficiently on GPU infrastructure using mixed-precision training, gradient accumulation/checkpointing, efficient attention, and distributed strategies (DDP, FSDP, or equivalent).
  4. Build and optimize feature pipelines on Databricks and Spark, transforming raw behavioral events into high-quality model inputs.
  5. Translate prototypes into production ML systems — scalable, reliable, and observable — and drive inference performance through architectural and serving-side optimization.

Skills

Required

  • Python
  • PyTorch
  • TensorFlow
  • end-to-end ML lifecycle
  • model optimization
  • inference efficiency
  • production system integration

Nice to have

  • fraud detection
  • anomaly detection
  • behavioral modeling
  • Adobe Experience Platform (AEP)

What the JD emphasized

  • build and develop machine learning models from scratch
  • manage the entire model lifecycle
  • actively building in-house behavioral foundation models
  • own deep learning systems end-to-end
  • 8+ years of professional experience building and deploying ML solutions at scale

Other signals

  • building and developing machine learning models from scratch
  • manage the entire model lifecycle
  • actively building in-house behavioral foundation models
  • own deep learning systems end-to-end