Responsibilities

• Design and build AI pipelines that handle real‑world messiness – long‑form media, unstructured documents, streaming data, multi‑modal inputs – including the chunking, context management, retry logic, and output validation that production systems require.
• Architect autonomous agent systems: tool selection, planning loops, memory strategies, evaluation, and the guardrails that keep them from going off the rails.
• Build on AWS using services like Bedrock, S3, Lambda, Step Functions, SageMaker, OpenSearch, and ECS/EKS – choosing the right primitive for the job rather than forcing one pattern everywhere.
• Make pragmatic decisions about model selection, prompt design, fine‑tuning vs. RAG vs. agentic workflows, and cost/latency/quality tradeoffs.
• Own the data layer: how data is ingested, transformed, stored, indexed, and surfaced to models so that outputs are accurate, traceable, and useful.
• Establish evaluation frameworks and observability so we know whether the system is actually getting better when we change it.Collaborate with engineers and stakeholders to translate fuzzy business problems into technical designs and shipped systems.

Responsibilities

• Design and build AI pipelines that handle real‑world messiness – long‑form media, unstructured documents, streaming data, multi‑modal inputs – including the chunking, context management, retry logic, and output validation that production systems require.
• Architect autonomous agent systems: tool selection, planning loops, memory strategies, evaluation, and the guardrails that keep them from going off the rails.
• Build on AWS using services like Bedrock, S3, Lambda, Step Functions, SageMaker, OpenSearch, and ECS/EKS – choosing the right primitive for the job rather than forcing one pattern everywhere.
• Make pragmatic decisions about model selection, prompt design, fine‑tuning vs. RAG vs. agentic workflows, and cost/latency/quality tradeoffs.
• Own the data layer: how data is ingested, transformed, stored, indexed, and surfaced to models so that outputs are accurate, traceable, and useful.
• Establish evaluation frameworks and observability so we know whether the system is actually getting better when we change it.Collaborate with engineers and stakeholders to translate fuzzy business problems into technical designs and shipped systems.