Selected projects spanning independent software systems and quantum-HPC infrastructure work.
A private assistant inspired by OpenClaw but built from scratch, connecting Telegram, OpenAI, and Google Calendar so scheduling and messaging happen from one conversational interface.
A digital-analog quantum SDK I originally led, focused on quantum machine learning, built on PyTorch, and designed to be fully differentiable for fast model-to-execution iteration.
A private assistant connecting Telegram, OpenAI, and Google Calendar endpoints in one flow. It was inspired by OpenClaw but built from scratch, and lets me schedule, reschedule, and trigger messages conversationally without switching tools.
A GPU-accelerated pipeline that detects events and extracts performance metrics directly from video streams. It cuts manual review time and makes longitudinal performance tracking much easier.
A travel-tech project where we used LLMs and map data to help people plan city holidays. The product was aimed at the hospitality sector so hotels and related operators could improve their guest offering while opening new revenue channels.
A proptech concept using AI models to support property due diligence and generate market ideas for individual investors. The pain point came from entering the market myself and seeing how fragmented and time-consuming investor research can be.
QPUs were often treated as special-case devices in HPC operations. I helped build QRMI and the Slurm/SPANK integration path so quantum hardware can be scheduled like other shared resources. This opened a practical route for CPU-GPU-QPU workflows in real environments.
To make high-level quantum code runnable in real infrastructure, I worked on Pasqal integration pathways connected to CUDA-Q. The result is a cleaner handoff from developer code to schedulable execution on heterogeneous systems.
CUDA-Q Docs Pasqal PR (Mar 20, 2025) Pasqal PR (Oct 28, 2025) ISC Talk
Deploying QPUs in HPC centers requires more than hardware installation. I led integrations across partner sites, including CEA and JSC, and shaped the orchestration model used inside allocations so sites can enforce priorities, access classes, and runtime policies. Together, these changes made QPUs operational infrastructure for day-to-day workloads, not isolated demonstrations.
I worked on execution paths spanning cloud QPUs, on-prem systems, and HPC-backed emulators so teams could use the same workflow across environments. That reduced handoff friction between development and production while enabling continuous validation when hardware access was limited. The result was faster iteration, fewer environment-specific surprises, and stronger reliability before production runs.
Hybrid Cloud Publication Emu-TN Publication IEEE QCE Talk OpenQSE Talk
I was the original lead of Qadence, a digital-analog quantum SDK focused on quantum machine learning and practical hybrid algorithm development. Qadence is built on PyTorch and was designed as a fully differentiable framework, making gradient-based workflows first-class and tightening the loop between model design, training, and execution. Beyond shipping open-source code, it helped set the software direction for how users move from exploration to repeatable execution in Pasqal-centered workflows.