Senior Data Scientist / ML Engineer with AWS

Robotics Technologies is hiring a Senior Data Scientist / ML Engineer to turn traveler and catalog data into narrative text and synthetic datasets, then use those materials to power large-scale deep learning and LLM pipelines on AWS. The role emphasizes distributed training, careful evaluation and error analysis, and production-focused inference that balances latency and cost.

What you’ll do

• Transform data into training sets: convert both unstructured and structured traveler and catalog data into narrative text documents and synthetic data at a scale of tens of millions of documents.
• Drive feature and context engineering: determine what information is included in each document and how it is expressed, including items like market-relative price, tier, and trip context.
• Lead large-scale distributed training: run distributed training on Amazon SageMaker AI, including HyperPod.
• Perform LLM fine-tuning: implement LoRA/QLoRA adapters and run supervised training and reinforcement learning approaches for ranking and reasoning.
• Build and tune deep learning models: design model architectures, develop ranking heads, and support co-training and multi-task setups.
• Deploy efficient inference: deliver low-latency, cost-efficient serving, including quantization, distillation, and profiling.
• Evaluate and analyze errors: create evaluation sets, run ablations, detect data leakage, and document results clearly for customers and the AWS science team.
• Handle data privacy and security: manage customer data safely, including personal-data removal (details continue in the privacy focus area).

Requirements

• Big data processing: Spark (PySpark) and the Hadoop ecosystem including Hive and HDFS for corpora of tens of millions of documents.
• AWS ML workflows: Amazon SageMaker AI processing, training, distributed training with HyperPod, and deployment.
• LLM training methods: LoRA/QLoRA fine-tuning and reinforcement learning approaches such as GRPO/PPO-style methods.
• Deep learning expertise in PyTorch: experience with architectures, training dynamics, and co-training.
• Production inference at scale: batch and production inference using vLLM and Amazon Bedrock batch.
• AWS privacy and security: personal-data removal, and controls including IAM, KMS, and Lake Formation.
• Clear communication: written communication with customer and science stakeholders.

Technologies you’ll work with

• Amazon SageMaker AI, HyperPod, LoRA, QLoRA, PyTorch
• Spark, PySpark, Hadoop ecosystem, Hive, HDFS
• vLLM, Amazon Bedrock, Amazon Bedrock batch
• IAM, KMS, Lake Formation
• Amazon Trainium, Inferentia, Neuron SDK
• GRPO, PPO, quantization, distillation, profiling, reinforcement learning

Preferred skills

• Experience with AWS Trainium / Inferentia via Neuron SDK.
• Recommender systems experience, including ranking metrics such as NDCG and MRR, plus sequential and generative recommenders.
• Experience with embedding quantization or semantic IDs.
• Background in a customer-facing or embedded role.

Location: Sanjose, CA (onsite).