I'm currently in the market for a job. My career has spanned a variety of domains, and I've developed expertise in many of them. This is well-illustrated by the work I've done over the past decade:

Unaffiliated • 2023—present

Most recently, I've been working on a book about software design. My approach is mostly grounded in linguistics and information theory, and I suspect it will be highly relevant to any attempt to use language models to create long-lasting, high-quality code. If you have a role that would let me put this theory into practice, I'm definitely interested in talking.

Microsoft • 2019—2023

I was a Principal Engineer at Semantic Machines, which had been acquired by Microsoft and placed within MSR. I was hired to help transform their prototype — a calendar assistant — into a general framework for conversational interfaces. My key contributions included:

• The prototype translated natural language utterances into a purpose-built programming language. I helped design and implement a from-scratch rewrite of that language.
• I proposed, designed, and implemented a sandboxed runtime for domain-specific functions, creating a clean separation between the framework and the applications built atop it.
• In natural language, objects are usually referenced indirectly, via predicates. In order to resolve these predicates into actual domain objects, we needed a constraint solver. Our prototype solver was, again, very calendar-specific. I designed and implemented a general solver, as described in this CACM article.

Unaffiliated • 2017—2019

I wrote a book about the Clojure programming language, and also noodled around with some personal projects. Some highlights:

• I implemented an efficient technique for rendering vector shapes on the GPU.
• While working on the rendering technique, I ran into a number of issues relating to floating-point imprecision in standard computational geometry techniques, which (I learned) assume you're using arbitrary-precision libraries. To mitigate this, I used some simple graph theory to detect and repair errors arising from this imprecision.
• I wrote a best-in-class implementation of functional data structures on the JVM.

Fitbit • 2015—2016

I was a Lead Engineer at Fitbit, which was in the process of splitting their monolithic core service into smaller services. The first step was to create a frontend service, which would authenticate and route each incoming request. They had written a prototype using Netty, and were looking for an expert in that library. Since I had been using Netty in my own open-source library for a number of years, I fit the bill. My key contributions included:

• I designed and implemented a from-scratch implementation of the frontend service.
• The core service was prone to instability, causing latency spikes. I designed and implemented a load-balancing algorithm to mitigate these issues, reducing p999 latencies by 50% on all API traffic.
• I trained my fellow backend engineers in the core principles of queueing theory.

Factual • 2012—2015

I was a Staff Engineer at Factual, which was a data provider for geographic locations. Its dataset was the original basis for both Facebook Places and Apple Maps, but pretty soon that customer base was exhausted. Predictably, for the time, they pivoted into mobile ad targeting. I joined shortly before the pivot, and wrote a lot of the software that sat at the interface between our data and the rest of the ad ecosystem. My key contributions included:

• I wrote and open-sourced libraries for large-scale data ingestion and a read-optimized key/value storage engine.
• I designed, implemented, and maintained services using these libraries.
• I founded and managed the San Francisco satellite office.