Specializing in Programming Languages (PL) under the supervision of Sam Tobin‑Hochstadt, minoring in Logic at the Department of Philosophy.

I work on fast run‑times of dynamically typed languages, particularly Racket. Initially, I investigated the effects of using the Racket byte‑code compiler as a front end for Pycket—studying optimizations like heavy inlining and lambda lifting and their effects on the Pycket run‑time.

When Racket decided to become more portable and started adopting Chez Scheme as its run time, I pivoted to help that effort. We had a rudimentary Racket implementation, Pycket, a meta‑tracing JIT compiler automatically generated by the RPython framework. It was designed to demonstrate that efficient sound gradual typing is possible. I helped make Racket more portable and adopt Chez Scheme fully:contentReference[oaicite:0]{index=0}. Mimicking Racket’s self‑hosting on Racket‑on‑Chez, I turned Pycket into a full implementation of Racket with a meta‑tracing JIT as its run‑time:contentReference[oaicite:1]{index=1}.

PyPy showed that fast prototyping of efficient run‑times for dynamic languages is possible through meta‑interpretation. My work takes this a step further to show that fast prototyping of efficient run‑times for self‑hosting dynamic languages is achievable:contentReference[oaicite:2]{index=2}.

Currently Pycket runs as fast as before on prior benchmarks, but self‑hosting introduces giant dispatch loops—similar to issues Mozilla encountered with TraceMonkey. You often lose more when slow than you gain when fast. My work suggests workarounds and shows that an efficient run‑time with a meta‑tracing JIT is still possible for fast prototyping of self‑hosting dynamic languages:contentReference[oaicite:3]{index=3}.

This program was a continuation of my MSc in natural language processing and machine learning:contentReference[oaicite:4]{index=4}. After another year of research on the system I built, I decided to pursue a PhD in programming languages instead.

I was part of the Text Analytics and Bio‑Informatics Lab. I studied the theory of computation and automated theorem proving, then focused on machine learning and natural language processing:contentReference[oaicite:5]{index=5}. I worked on a closed‑domain question‑answering system for Turkish high‑school students (a government‑funded project).

I gained both theoretical and practical experience. I programmed in imperative, functional and logic languages, focusing on the theory of computation and programming languages design and implementation:contentReference[oaicite:7]{index=7}.

I work in the Juju core team of about 15 people globally distributed. I help develop and maintain Juju’s client packages and libraries, including the CLI and python‑libjuju, and the Terraform Juju Provider. I’m involved in hiring, code reviews and documentation work.

Taught Introduction to Computer Science (HtDP‑style functional programming) using Racket.

I taught Advanced Compilers, Introduction to Programming (in Racket), and domain‑specific language design and implementation with Sam Tobin‑Hochstadt for one semester. I was his research assistant for five semesters and four summers.

Worked on developing the back end of virtual point‑of‑sale applications between merchants and banks. Technologies: Java, Spring, Tomcat, Mercurial, Jira.

Developed a Google Glass application (on a mock API) that uses CitySDK data to provide smart‑city tasks such as finding a suitable bus route.

I created an extra‑curricular game for students to “level up” their computer science skills. Each level required learning concepts and solving problems under constraints. I led a team of undergraduates as they leveled up.

Taught six semesters: Data Structures & Algorithms (Java), Principles of Programming Languages (Racket) and Database Systems with Data Mining.

Taught Lambda Calculus to math graduate students and “Programming = Algebra” to high‑school students with Chris Stephenson.

Reviewed two papers on statistical information retrieval for SIU 2014 and a paper on question‑answering systems for COLING 2014.

Led the preparation and compilation of the book of abstracts and proceedings for the conference.

Juju is an orchestration and modeling tool for deploying complex cloud workloads. It manages charmed operators for Kubernetes and traditional Linux apps with declarative relations for automated microservice integration. It also facilitates day‑2 operations such as scaling, upgrading and monitoring. As part of the core team, I help maintain the clients (CLI and python‑libjuju) and the facade API, integrate charmhub, design data schemas for migrating from MongoDB to dqlite, and handle general bug fixes and maintenance. I mostly use Go and Python.

The Terraform Juju Provider enables integration with Juju while managing Terraform environments. I started developing it as part of my work on the Juju API. I implemented new features, migrated from SDK 2 to the new Terraform provider framework, and maintained the release cadence for new versions. All in Go.

A meta‑tracing JIT originally designed to reduce the overhead of sound gradual typing, now a run‑time for self‑hosting full Racket. I am the main developer and maintainer. I re‑implemented the front end to make Racket bootstrap itself on Pycket, achieving a full Racket implementation on the RPython framework. I developed performance‑debugging tools, implemented various optimizations and many run‑time primitives, and maintained data‑structure implementations (hash tables, vectors, structs, etc.). For details, see the PhD description.

[Pycket repo]

A full‑stack nanopass compiler from a small subset of Racket down to x86‑64 assembly. It started for fun, then became a school project and later an open‑source project. I built and tested all the passes (closure conversion, register allocation, code generation, etc.).

[Rax repo]

I collaborated with Jeremy Siek and Matthew Heimerdinger on proving the correctness of a closure‑conversion pass of a compiler for a functional language represented via denotational semantics in Isabelle.

[denotsem-compiler repo] (may be private)

A closed‑domain question‑answering system for high‑school students funded by TÜBİTAK. I led research and development with three faculty members (CS and School of Education) and four graduate students. It was my MSc thesis work, so I handled nearly every aspect (see Publications). I developed a hybrid model combining Hidden Markov Models and rule‑based methods for question analysis, a reliability score for web documents, and an IR module using the Indri search engine for answer generation. I implemented both the front and back ends in Python.

Reading is a kind of foreign language for Deaf people. I collaborated with the Department of Western Languages & Literature to build a web‑based application to teach people with hearing disabilities how to read. Two linguists, one CS faculty and two research assistants were involved. I helped implement the application (for PC and mobile) designed specifically for Deaf users, using JavaScript, HTML and PhoneGap.

A data‑collection, processing and visualization framework for real‑time social‑network data led by Suzan Üsküdarlı. With four faculty members and seven grad students, I developed parts of the processing library in JavaEE, implementing tools for analyses such as centrality and stop‑word collection.

[Direnaj staging site] (may be private due to Twitter privacy policies)

A resolution‑refutation automated theorem prover I wrote for fun and later turned into a project. Written entirely in Racket.

[FARS repo]

My undergraduate thesis (see BSc description) was part of a project at Brown University supervised by Shriram Krishnamurthi, implementing a compiler from Racket’s Advanced Student Language to JavaScript and HTML. I worked on the compiler and run‑time libraries; my major contribution was adding tail calls. The work is now used in WeScheme for the Bootstrap curriculum.

[moby-scheme repo]