Features & Labels

Head of Engineering • 2024 — Now

Founding Engineer • 2022 — 2024

Built the serverless GPU runtime that is powering of thousands of customers, as well as the underlying platform. Led the efforts for building the core IP -- the most efficient inference engine for diffusion models. Scaled the team from 5 people to 13 people.


Software Engineer • 2020 — 2021

Helped building tools to improve workflow of MLOps scenarios, dealt with large data transfer optimizations, and was primarily responsible for formalizing the inter-filesystem operations.


Python Specialist • 2020

Worked on the static analysis engine to improve platform's support and semantical understanding of Python. Built analysis rules to run over millions of LoC and find potential bugs in the open source ecosystem. Was also responsible for the Python parser support and auxiliary tooling for automatic rule generation.

F/OSS Projects

Actively contributing since 2018 (and a committer since 2020). Mainly specializing in the parser and the bytecode compiler subsystems (as well as maintaining the AST module). Some of the more public features I've worked on include:

Co-maintainer of the project; initially implemented Black Gallery, a tool for previewing possible changes in popular open source projects before releases to ensure stability. Currently maintaining black's underlying CST implementation.

Involved in the design of the specification, as well as the implementation of certain features (like the new callbacks API), and creation of new filesystem adapters (which includes one of the fastest SFTP clients for Python). Also lots of contributions to the existing implementations of the different cloud providers (ranging from optimizations to new features).

Co-maintainer of the project, primarily worked on syntax errors as well as the 3.9/3.10 migration. Also experimented with various techniques to adapt the existing LL(1) parser with error recovery.

Authored Refactor, a popular library for refactoring Python source code directly at the AST level.

Worked on the PyPy 3.8 with implementing various features like typed_ast and f-string debugging expressions, as well overall fixes on the language compliance. Had hands-on experience with RPython JIT.

Authored Reiz, a source code search engine to express fragmental knowledge and gather data about source code (as well as to sample it). Involved building a custom IR, designing multi-layered DSLs (ReizQL for the engine itself, and IRun as a Python-flavor for easily expressing python syntax for queries) and compilers for both of them.

Started Teyit for automatically improving the quality of Python unittests. It automatically detects and upgrades the assertions.

Publications & Given Talks