Tools, benchmarks and products I’ve built.
From research prototypes to products used by real teams: fuzzing engines, benchmarks for AI agents and program analysis tools. Here are a few of my favorites.

FuzzSlice
Cut Through the Noise! Target Actionable Vulnerabilities in your static analysis reports. FuzzSlice employs slice-level fuzzing which we presented at ICSE 2024 to efficiently fuzz your codebase.
Website

Mayil
Solve tickets faster. Without changing the way you work! Mayil brings years of Bug Localization research to help your engineers get started faster
Website
AssertFlip
An agentic approach to bug reproduction: it has an LLM write a passing test from the bug report, then flips it into one that fails on the bug. When released, it outperformed all existing bug-reproduction techniques.
Repository

BouncerBench
Sometimes no answer is better than a wrong one. A benchmark that checks whether LLM-based software agents know when to abstain from ambiguous or unsolvable tickets.
Repository

IBM/tree-sitter-codeviews
Extract, combine and create custom source code-views for your Java/C# projects using tree-sitter! (pypi/comex)
Repository
Pydevaa
Pydevaa is a framework that automatically detects and attempts to exploit vulnerabilities (XSS, Fragment Injection, Intent Reflection) in Android applications. Make a pull to add more!
Repository
PillaiPidi
A tool to perform cross-language static data flow analysis. Find vulnerabilities across the JNI bridge!
Jar Package

depend
Transitive dependency chain analyzer (Supports Python, JS, C#, PHP, Rust & GO Projects). This project is now an internal scanner to DeepSource.
pypi/depend
mocktail
A streamlined pipeline to create source code representations for ML tasks that capture syntactic and semantic program features by utilizing multiple intermediate representations of code.
Repository
AiR
An Android app that visualises air pollution around you in augmented reality. Featured by the UN Environment Programme; presented at IEEE VIS 2021.
Paper