AI Application Engineer Intern

What You’ll Learn

• Project Overview: Network-attached direct storage (NADS)
• Skills You’ll Learn:
  • Deploying and tuning LLM inference serving (vLLM, KV cache offload with LMCache/NVIDIA Dynamo) on data-center GPUs
  • Benchmarking AI system performance end to end, using latency and throughput metrics such as time-to-first-token and goodput

What You’ll Do

The SoC Architect team proposes multiple SoC level architecture candidates for emerging memory/storage solution and demonstrates that the proposed architecture is viable through hands-on RTL implementation.

This role offers the opportunity to shape next-generation SoC and subsystem architectures while staying deeply hands-on in RTL design, FPGA integration, and validation. It is ideal for engineers who want to turn advanced memory and storage concepts into real hardware and drive them all the way to silicon.

Location: Daily onsite presence at our San Jose, CA headquarters in alignment with our Flexible Work policy

Reports to: Senior Director

• Deploy and operate an LLM inference service (vLLM serving Llama-3.3-70B) on a 4 × NVIDIA H200 GPU server, working closely with your mentor and the storage architecture team.
• Build an automated benchmark harness that replays long-context, multi-turn and agent-style request traces and records latency and throughput (TTFT, TPOT, goodput).
• Configure KV cache offloading (LMCache, NVIDIA Dynamo/NIXL) to network-attached storage over NVMe-oF and RDMA, first on a BlueField-3 storage server and then on our network-attached SSD prototype.
• Run daily experiment sweeps across request rates, context lengths and cache sizes, keeping every configuration and result reproducible in git.
• Analyze results with Python to find bottlenecks across the GPU, network and SSD, and share findings in weekly team meetings.
• Collaborate with FPGA, SoC and firmware engineers to debug the end-to-end data path and validate KV cache correctness.
• Summarize the measured storage requirements (bandwidth, IOPS, latency) that shape the architecture of future SSD controllers, and present them to the team at the end of the internship.
  
• Complete other responsibilities as assigned.

What You Bring

• Pursuing Master's or PhD (preferred) in Computer Science, Computer Engineering, Electrical Engineering, or a related field preferred.
• Must have at least 1 academic quarter/semester remaining.
• Hands-on experience running LLM inference with Python and PyTorch on NVIDIA GPUs under Linux, using a serving framework such as vLLM, SGLang or Hugging Face TGI.
• Understanding of LLM serving internals: KV cache, prefix caching, batching and parallelism (tensor/pipeline), and latency metrics such as TTFT, TPOT and throughput.
• Experience benchmarking and profiling system performance, and analyzing results with scripts (e.g., Python, pandas, plotting) to explain bottlenecks.
• Familiarity with KV cache offloading (LMCache, NVIDIA Dynamo/NIXL), RDMA/RoCE or NVMe/NVMe-oF storage, and containers (Docker/Kubernetes) preferred.
  
• You’re inclusive, adapting your style to the situation and diverse global norms of our people.
• An avid learner, you approach challenges with curiosity and resilience, seeking data to help build understanding.
• You’re collaborative, building relationships, humbly offering support and openly welcoming approaches.
• Innovative and creative, you proactively explore new ideas and adapt quickly to change.

#LI-AD1

Hourly Base Pay Rate 28 - 68 USD