About The Role

At JPMorganChase, we are building the infrastructure that powers the next generation of enterprise AI — and we need talented engineers who are passionate about LLM inference to help us do it. This is your opportunity to work at the intersection of cutting-edge machine learning and large-scale production systems, directly contributing to how one of the world's largest financial institutions deploys and optimizes AI at scale.

As a Lead Software Engineer at JPMorganChase within the AI/ML Data Platform team, you will be a key technical contributor on LLM inference performance — supporting optimization strategy, benchmarking, and efficiency at scale. You will collaborate closely with senior engineers and engineering leadership to help shape how our platform evolves, ensuring every model we serve is fast, cost-efficient, and production-ready. This is a high-impact individual contributor role where your technical contributions will have direct, measurable influence on the firm's AI capabilities.

Job Responsibilities

Execute systematic benchmarking and performance characterization across production LLM workloads, establishing reproducible baselines, identifying regressions, and quantifying the impact of configuration changes before they reach production
Design and run quantization experiments — FP8, INT8/INT4 (GPTQ/AWQ), and next-generation precision formats — measuring accuracy delta, throughput improvement, memory reduction, and cost-per-token impact
Support speculative decoding strategy across the model portfolio, including draft model, n-gram, and multi-token prediction approaches, contributing to acceptance rate measurement and per-workload configuration recommendations
Build and maintain GPU efficiency metrics covering utilization, memory headroom, cost per 1K tokens, and waste identification — providing engineering teams with a data-driven view of platform efficiency
Benchmark the platform against external providers and published industry numbers, identifying gaps and contributing to improvement initiatives
Participate in inference engine upgrade evaluations, including new scheduler architectures, async tensor parallelism, disaggregated prefill/decode, and advanced speculative decoding, supporting systematic validation before production promotion
Contribute to GPU chaos engineering efforts, including induced failure scenarios, hardware diagnostic monitoring, and detection and recovery measurement
Leverage enterprise-authorized AI coding assist tools within the work environment to improve code quality, delivery speed, and productivity (e.g., code generation/refactoring, unit test creation, documentation), while validating outputs through peer review, automated testing, and secure coding standards
Apply knowledge of tools within the Software Development Life Cycle toolchain, including enterprise-authorized AI-assisted development and automation capabilities, to improve the value realized by automation

Required qualifications, capabilities, and skills

Formal training or certification on software engineering concepts and advanced applied experience – preferably Go / Python
Hands-on experience with LLM inference systems — vLLM, TensorRT-LLM, SGLang, LLM-D, or equivalent production serving engines
Strong understanding of GPU memory architecture, including KV cache sizing and dynamics, memory-bandwidth versus compute bottlenecks, and the practical implications of quantization at inference time
Experience with quantization techniques and their real-world tradeoffs at scale
Familiarity with speculative decoding and the variables that drive acceptance rates in production workloads
Rigorous benchmarking skills using GuideLLM, custom harnesses, or equivalent tooling, with the ability to support every performance claim with data
Experience operating in cloud GPU infrastructure at scale (AWS, Kubernetes-based managed inference services)
Ability to communicate technical trade-offs clearly to engineering peers and senior stakeholders
Hands-on experience using enterprise-authorized AI-assisted software development tools within the work environment (e.g., for coding, testing, troubleshooting, or documentation) with demonstrated ability to critically evaluate and validate AI-generated outputs
Understanding of responsible AI use in engineering workflows, including data sensitivity considerations, secure handling of inputs/outputs, and adherence to resiliency and security expectations

Preferred qualifications, capabilities, and skills

Experience with disaggregated prefill/decode serving architectures
Familiarity with GPU hardware diagnostics tools such as DCGM, NVML, or XID event tracking
Experience with ML observability and production monitoring for inference workloads
Awareness of the LLM inference competitive landscape with a track record of applying industry benchmarks to drive platform improvements