Principal Software Engineer - AI Foundations

As a Principal Software Engineer at JPMorganChase within the Chief Data and Analytics Office (CDAO), you provide expertise and engineering excellence as an integral part of an agile team to enhance, build, and deliver trusted market-leading technology products in a secure, stable, and scalable way.

In this role, you will lead the design and evolution of the firm's GenAI serving platform, focused on high-performance LLM inference, intelligent model routing, and GPU efficiency, to deliver reliable, cost-effective AI capabilities at enterprise scale.

Leveraging your advanced technical capabilities and collaborating with colleagues across the organization you will drive best-in-class outcomes across various technologies to support one or more of the firm's portfolios.

Influence leaders and senior stakeholders across business, product, and technology to drive alignment and outcomes; foster a culture of diversity, opportunity, inclusion, and respect.

Job Responsibilities


* Design, build, and operate a high-throughput, low-latency LLM serving platform (batching, scheduling, caching, streaming responses, multi-tenancy, and autoscaling) across GPU/CPU fleets.


* Build and evolve a GenAI Gateway / inference API layer (authentication, authorization, quota/rate limiting, routing, request shaping, policy enforcement hooks, and standardized observability) to support diverse application workloads.


* Develop and optimize "open routing" / intelligent model routing across multiple model backends (open-source and vendor models), balancing quality, latency, reliability, and cost with configurable policies and guardrails.


* Drive GPU serving optimization: kernel-level performance tuning where needed; model compilation/acceleration (e.g., TensorRT-style approaches), efficient memory management, KV-cache strategies, and throughput tuning (prefill vs.

decode optimization).


* Implement quantization and compression strategies (e.g., INT8/INT4, weight-only quantization), including evaluation-driven selection and safe rollout practices that preserve quality and reduce cost/latency.


* Design and implement disaggregated serving patterns (e.g., separating prefill/decode, KV-cache offload, tiered serving) and distributed inference architectures to improve utilization and tail latency.


* Develop secure, high-quality production code; review, debug, and improve code written by others; create durable, reusable frameworks and platform components leveraged across teams, aligned to modern product development methodologies.


* Own and support SDK and service integrations, ensuring reliability, performance, and maintainability.


* Establish SLOs/SLAs for inference services and build operational excellence (load testing, capacity planning, incident response playbooks, regression detection, and continuous performance benchmarking); build robust performance and cost observability (latency histograms, token throughput, GPU utilization, memory f...