Experience

• Lead Frontiers, spanning inference and chip architecture for transformer inference ASICs, from die and package through racks to the serving stack
• Translate customer workloads, including frontier LLMs, large mixture-of-experts models, and long-context and agentic serving, into product requirements, and own the resulting tradeoffs across compute density, memory hierarchy and bandwidth, scale-up interconnect, numerics, power delivery, and thermals
• Build system-level performance and cost models covering chip throughput and latency, rack power, and total cost of ownership; benchmark against GPU baselines and field technical diligence from customers and analysts
• Sit in architecture reviews and software working groups across firmware, emulation and simulation infrastructure, and pre-silicon validation, tracking blockers and driving timelines toward bring-up
• Drive the customer-facing inference software stack, from serving frameworks such as vLLM and SGLang to the runtime interfaces customers deploy in production
• Lead technical evaluation and commercial engagement with suppliers across silicon, memory, packaging, and interconnect, and hire for architecture and inference engineering teams

• Partnered with business unit heads and senior functional leaders on strategic initiatives across trading, research, technology, and operations
• Built performance analyses and operating metrics used in executive decisions across the firm

• Sized the market and drafted the go-to-market strategy for a new enterprise cybersecurity offering
• Identified cost opportunities in a medtech company's SG&A function and partnered with senior leaders on the execution roadmap
• Evaluated a private equity investment in the pharmaceutical sector through customer interviews, competitive analysis, and market research

Languages & frameworks

Domains

• Concurrent master's and bachelor's degrees
• Teaching Fellow for Statistics 110: Introduction to Probability
• Projects in matrix completion for image reconstruction, parallel reinforcement learning, and generative adversarial networks