Creations Born from the Cosmic Wave

Over the years, I’ve worked on numerous projects. While some of my projects are proprietary, I also have open-source work and licensed projects that I’m excited to share. Let me know if anything piques your interest!

  • Segmnts: AI-Native Collaboration for the Future of Work

    As Co-Founder & CEO at Segmnts, I spearheaded the development of a unified, AI-native platform that reimagines traditional office tools. Unlike conventional suites, Segmnts integrates AI agents that seamlessly modify, transform, and interconnect tools like documents, spreadsheets, and presentations. We introduced Pins, a novel agentic system that enables users to create and automate workflows with reusable AI-driven actions. This approach empowers teams to collaborate with AI coworkers, redefining productivity and setting a new standard for intelligent workplace automation.

    segmnts.com

  • Scalable Bid Management System for Google Shopping

    I led the development of a large-scale bid management system for Google Shopping, tackling complex regression problems involving feature engineering, time-series analysis, and ML Ops. As a Lead Data Scientist, I ensured the system's strategic and technical soundness, while as a Lead Machine Learning Engineer, I designed the ML pipeline's integration into broader systems. The project introduced robust quality gates for daily prediction monitoring, improving bidding efficiency and business performance. Our work enhanced automation and adaptability in ad bidding, setting new benchmarks for AI-driven ad optimization.

    smarter-ecommerce.com

  • QA Automation for Games, Epoch ML

    At Epoch, I spent two years leading the research, design, and implementation of automated Game QA solutions, including capture-and-replay, localization testing, and AI-driven testing using imitation learning. I developed core testing features that enabled unit tests for gameplay, text scripts, and UX. As the company shifted from research to product, I transitioned into full-stack development, delivering user-facing features like video capture, collaborative tools, and a custom dashboard infrastructure. My work helped shape a scalable, automated QA platform, improving efficiency for game studios and setting a new standard for AI-driven game testing.

    epochml.com

  • Quad Meshing for Implicit Surface-Based Game Engine

    At Unbound.io, I led the research and development of a meshing pipeline for implicit surfaces, enabling their use in external tools like Blender. Implicit surfaces provide powerful modeling capabilities but require precise meshing to avoid artifacts and maintain visual fidelity, similar to digitizing an audio signal. As Lead Researcher, I designed and implemented a quad meshing solution that ensured regular, grid-like topology while integrating seamlessly into the game engine. This work improved the adaptability of implicit modeling for game development, expanding its usability beyond proprietary environments.

    unbound.io

  • Semantic Clone Detection Benchmark

    As a researcher, I developed a comprehensive benchmark for evaluating clone detection tools, specifically their ability to identify semantic clones with 0% syntactic similarity. This benchmark provides a structured dataset and rigorous evaluation framework, enabling objective performance assessment and fair comparisons between detection methods. By standardizing evaluation criteria, it advances research in semantic code analysis, improving the accuracy and reliability of clone detection techniques.

    github.com/hannes-thaller/semantic-clone-detection-benchmark

  • Gradient: Advancing Software Analysis with Probabilistic Software Modeling

    In my role as a researcher and speaker, I led efforts in Probabilistic Software Modeling (PSM), developing techniques that transform software into probabilistic models for enhanced fault detection, semantic analysis, and predictive insights. Our work introduced novel machine learning-driven methods to analyze software structure and behavior, improving automated debugging and anomaly detection. This approach has significant industry implications, enabling more scalable and intelligent application performance management, reducing maintenance costs, and enhancing software reliability in critical systems like finance, healthcare, and embedded software.

    github.com/hannes-thaller/gradient

  • Pillar: Detecting Design Patterns with Feature Maps

    As Researcher and Developer, I pioneered a novel approach to program analysis by creating feature maps—visual representations of a program's structural properties. Using static code analysis and graph transformations, we projected key program features into matrix-based images, enabling automated design pattern detection. This innovation provided a new way to analyze software architecture, improving code quality assessment and advancing automated software engineering methodologies.

    github.com/hannes-thaller/pillar

  • ECCO: Feature-Oriented and Distributed Version Control System

    As a developer, I contributed to the early ECCO UI, integrating it into Eclipse to manage software versions and variants efficiently. ECCO is a feature-oriented, distributed version control system that supports variability across different artifacts. My work improved feature location, product line engineering, and automated reuse, enhancing scalability and efficiency in software development. This innovation has industry-wide impact, optimizing variant management and distributed development.

    github.com/jku-isse/ecco