Senior Technical Program Manager

Parallel Domain welcomes Zack Novak as CEO. Parallel Domain bet the company, and it paid off. A little more than a year ago, Parallel Domain made a defining decision. Parallel Domain walked away from the procedural environment creation approach Parallel Domain had spent years building and shifted its entire focus to a scalable reconstruction method. It was a genuine bet-the-company move. The result was PD Replica. It wasn't just a new product. It represented a fundamentally different way to simulate the real world. PD Replica turned out to be both more scalable and more performant, enabling its customers to test and validate their autonomous systems faster, across more scenarios, and with greater fidelity than ever before. It fueled a year of massive growth across testing and validation use cases. That conviction is now the engine behind everything Parallel Domain is doing. The market is shifting in its direction. Something fundamental is happening in autonomy right now. The industry's shift from modular perception-planning-control stacks toward end-to-end learned models is driving significantly greater demand for high-fidelity simulation data. When your entire autonomy stack is learned, the quality and diversity of your simulation environment becomes the bottleneck, and that's exactly the problem PD Replica was built to solve. This isn't just its view. Public presentations from Waymo, Tesla, and Waabi have confirmed the central importance of simulation and 3D scene reconstruction for achieving production-ready autonomous deployment. The validation is real, and it's accelerating demand across the board. Its core autonomous vehicle customers are pulling harder than ever, but the opportunity no longer stops at passenger cars. Over the past year, Parallel Domain has seen real adoption from companies building drones, eVTOL aircraft, last-mile delivery robots, and agricultural autonomy systems. The underlying challenge they all share, rigorously testing autonomous systems at scale before they ever touch the real world, is universal. Why Parallel Domain is bringing on a CEO now. Kevin and its Board looked at the progress the team has made, the traction PD Replica has earned, and the scale of the market ahead, and recognized that the moment demands dedicated operational and commercial leadership to scale the business. Zack Novak joins Parallel Domain as CEO to build the operational and commercial engine the company now requires; scaling not just go-to-market, but the organization as a whole. Zack has spent his career scaling industrial AI and enterprise software businesses, most recently at Uptake and Quantix, where he drove go-to-market execution in complex, technical markets. His experience is precisely what Parallel Domain need to capture the opportunity across automotive and the new industries pulling Parallel Domain forward. Kevin goes deeper, not sideways. Kevin McNamara isn't stepping back. He's going deeper, moving fully into product, engineering, and working directly with customers to accelerate the value PD Replica delivers. It's also where the company needs him most. The technology landscape in reconstruction, simulation, and AI is evolving rapidly. Parallel Domain fully expect Parallel Domain'll need to reinvent aspects of its approach again as the field advances. That kind of foresight requires a founder who is close to the product and close to customers, not one managing a P&L from a distance. Kevin will spend his time with engineering, product, and the people who use PD Replica every day, making sure Parallel Domain stays at the frontier. A partnership built to scale. Zack drives go-to-market and commercial execution. Kevin stays embedded in the technology and with customers. Together, they form a partnership designed to let Parallel Domain do two things at once: scale aggressively into a rapidly expanding market and continue to lead on the product and technology that got Parallel Domain here. Welcome to Parallel Domain, Zack.

From sparse data to photorealism: accelerating Physical AI with Parallel Domain and NVIDIA Fixer. In the development of Physical AI, the ability to trust your simulation is everything. At Parallel Domain, Parallel Domain reconstruct reality from sensor data, powering photorealistic simulation to test and verify end-to-end model and perception performance. To verify autonomous systems effectively, Parallel Domain need environments that are more than just visual approximations, they must be geometrically accurate and capable of withstanding rigorous testing. That is why Parallel Domain is excited to unveil its latest technical advancement at CES: the integration of NVIDIA Omniverse NuRec Fixer into its PD Replica pipeline. The challenge: the gaps in reality. Building a digital twin from real-world data is complex. Sensors have limitations, occlusions occur, and capturing every angle of a dynamic environment is often impossible. Traditional reconstruction methods can struggle with these gaps, leading to visual artifacts or inaccurate geometry when rendering views that weren't explicitly captured by the original sensors. When a simulation platform relies on messy input data, it limits the developer's ability to test rigorously. The solution: closing the gaps with NVIDIA Omniverse NuRec Fixer. NVIDIA Omniverse NuRec Fixer is a diffusion-based model built on the NVIDIA Cosmos Predict world foundation model that removes rendering artifacts and restores detail in under-constrained regions of a scene. By incorporating Fixer into the PD Replica creation workflow, running on NVIDIA GPUs, Parallel Domain is transforming how messy real-world data is converted into simulation-ready assets. The results Parallel Domain is seeing in its internal testing are transformative. By utilizing Fixer, its pipeline can now: * Master Novel Poses: One of the hardest challenges in simulation is rendering "off-axis" views, angles the original capture never saw. Fixer excels here, allowing Parallel Domain to balance real vs. virtual viewpoints and generate valid data for large off-axis views. * Clean Up Artifacts: Where base reconstruction sometimes leaves noise, Fixer acts as an intelligent post-processor. It significantly reduces artifacts and mitigates shifts in color space, resulting in a cleaner, more consistent image. Visual proof. Left - PD Replica novel view output. Right - PD Replica + NVIDIA Fixer novel view output. In early comparisons, the difference is stark. While the Base Replica provides a solid foundation, the Replica + Fixer output sharpens dynamic objects and smooths out noise, creating a scene that is simulation-ready. This improvement in geometric accuracy allows perception teams to trust that an obstacle in the simulation will trigger the same response as it would in the physical world. Powering Software Augmented Testing. This integration reinforces its commitment to Software Augmented Testing. By combining the high-fidelity reconstruction of PD Replica with the neural rendering capabilities of Fixer, Parallel Domain enable developers to: * Test the Untestable: Generate scenarios that are too dangerous or rare to hunt for on the road. * Verify Perception Performance: Use pixel-perfect ground truth to run regression tests on perception stacks with confidence. * Scale Without Overhead: Achieve these high-fidelity results without the massive data requirements of purely generative AI approaches. Collaborating to drive innovation. Parallel Domain has been working closely with NVIDIA to push its tech forward. By leveraging NVIDIA AI infrastructure and the Fixer architecture, Parallel Domain is helping simulation power physical AI development more broadly. Together, Parallel Domain is turning sparse, noisy real-world data into the pristine, annotated environments required to deploy safe autonomous systems. Visit Parallel Domain at CES to see how Parallel Domain is defining the future of sensor simulation and physical AI development with NVIDIA technology.