Apr 2022 – Apr 2026 · 4 yrs · San Francisco Bay Area · Remote

• Lead a team of 3 engineers in analyzing all performance aspects of the Airbnb site.
• Responsible for the performance roadmap aligning with all partner organizations
• Work with the cloud team to scale their Kubernetes fleet
• Analyze and optimize Java applications using JFR data and performance profiles
• Selection of optimal AWS hardware to maximize performance and cost including compute and GPUs
• Owner of AWS cost reduction efforts including EC2, S3, EBS, ElastiCache ECR etc.
• Lead the effort to enable Arm64 hardware in the Airbnb fleet.
• Work with the machine learning infra team to optimize model serving and training.
• Responsible for the Kubernetes container runtime.
• Work extensively with the Search team to optimize their Tensorflow ranking system

Mar 2021 – Sep 2021 · 6 mos · San Francisco Bay Area

• Analysis of performance on all Snowflake tiers, Cloud Services (Java), Query Processing (C++) and FDB (C++).
• Creation of numerous tools to collect performance data at all levels. Analysis of query processing using bpf leading to the discovery of a number of performance bottlenecks. Analysis of new FDB features using both perf tool and bpf to generate flamegraphs and highlight resource usage.
• Rebuilt and optimized the query processing code with Clang / LLVM compiler.
• Applied profile feedback and BOLT using a locally built compiler. Used TPC-H queries for feedback.
• Contributed to a cross functional team analyzing TPC-H performance on Snowflake, deep dive analysis of several compute bound queries.
• Assessed cpu, memory and flash performance on next gen hardware from cloud providers. Partnered closely with Intel to analyze query processing using AWS bare metal instances and hardware performance counters using emon.
• Created new Azure images to enable their Gen 2 hypervisor to improve performance.

Jan 2015 – Mar 2021 · 6 yrs 2 mos · Menlo Park

• One of the most senior performance engineers at FaceBook.
• Developed PerfDoctor, Facebook’s premier performance tool which analyzes all aspects of the performance of a service at scale, extracting data, detecting performance issues, highlighting fixes. Evolved and enhanced PerfDoctor for Machine Learning training systems. Developed over 100 performance detectors that run daily on all major services and flag performance issues for developers.
• Analyzed the performance of hundreds of services, designed and implemented performance fixes. Partner closely with software teams to drive performance optimizations into production.
• Increase the scalability of many services through workload characterization and targeted optimizations.
• Teach two day intensive course covering all aspects of performance from kernel to applications.

Jul 2013 – Jan 2015 · 1 yr 6 mos · San Francisco Bay Area

Sep 2012 – Jul 2013 · 10 mos

• Drive performance and monitoring features in upcoming MarkLogic releases. Expert in MarkLogic deployment to the Amazon EC2. Work with partners such as StorageReview.com and Cloudsoft to improve MarkLogics public profile. Analyzing performance for Customer Proof of Concepts. Presenting Performance and EC2 at MarkLogic World

Jun 2011 – Sep 2012 · 1 yr 3 mos · Los Gatos, CA

• Senior member of a team which architected the performance of the Netflix site on Amazon’s EC2 cloud. Architect of a highly scalable Cloud Performance framework using Jenkins Continuous Integration and JMeter to stress RESTful, Java and Cassandra based services. Cassandra performance expert doing deep dive analysis of scalability on EC2. Joint author of Netflix technical Blogs, Benchmarking Cassandra Scalability on AWS – Over a million writes per second and JMeter Plugin for Cassandra

Mar 2010 – Jun 2011 · 1 yr 3 mos · Redwood Shores, CA

• Developing the interface between Oracle's flagship RDBMS and all major OS variants with focus on Linux (64-bit), Solaris (SPARC and x86) and HPUX

Apr 2007 – Feb 2011 · 3 yrs 10 mos

• Member of the team defining the next generation of SPARC volume processors. Define architecture enhancements to accelerate Virtualization, hardware threading, Java, crypto and networking. Also involved in defining the performance counter framework for the processors.
• Member of a virtual team deploying and enhancing open source Web 2.0 software stacks including Memcache and
• Hadoop on Sun's CMT servers
• Performance evaluation and software enhancements for UltraSparc T1 and T2 based systems, from chip bring up through delivery and launch.

Jan 1994 – Jan 2007 · 13 yrs

Jan 1993 – Jan 2007 · 14 yrs

Jan 1992 – Jan 1994 · 2 yrs

Jan 1986 – Jan 1989 · 3 yrs