How NorthStar Medical Radioisotopes Cut Simulations from 4 Days to Under an Hour

The growing demand for AI and advanced simulation tools for nuclear medicine research presents significant challenges in scalability and research efficiency. Hoonify partnered with NorthStar to deliver the TurbOS® software platform, which included a complete suite of performance-optimized research tools that produced faster Monte Carlo N-Particle (MCNP) simulation results. To eliminate support challenges and reduce costs, Hoonify delivered TurbOS as a turn-key, private cloud solution which empowered NorthStar engineers to perform their own complex simulations as well as reduce costs from extra subscriptions and technical assistance. The net result was increased research productivity and faster results.

NorthStar Medical Radioisotopes used to wait four days for a single MCNP simulation. On TurbOS®, the same simulation finishes in under an hour. Here’s how Hoonify moved NorthStar from aging on-prem hardware to a private cloud built for MCNP, and went live in one day.

The results at a glance

• 96× faster MCNP simulations, from 4 days to under 1 hour
• 33% increase in research throughput
• 50% cost savings compared with standard public cloud
• One day from legacy on-prem to live on the private cloud
• Simulation work rose from 60% to 80% of engineer time, with IT involvement down 20%

About NorthStar Medical Radioisotopes

NorthStar Medical Radioisotopes develops, produces, and commercializes radiopharmaceuticals for therapeutic and medical imaging applications. The company is an emerging leader in U.S. medical radioisotope research, and its development work depends heavily on computational modeling, especially Monte Carlo N-Particle (MCNP) simulations.

Why Hoonify hosts MCNP

MCNP is a trusted application for nuclear simulation research, and it runs best on high performance computers that can spread the work across many processors in parallel. Access to MCNP is tightly controlled. Oak Ridge National Laboratory’s Radiation Safety Information Computational Center (RSICC) endorsed Hoonify to host secure MCNP access for select, authorized users.

RSICC’s security and data-handling requirements call for stronger cyber security protections than most cloud providers offer. TurbOS provides them. That’s why Hoonify is one of the only private server operators endorsed by ORNL’s RSICC.

The challenge: multi-day simulations were the bottleneck

NorthStar needed faster, repeatable MCNP results to keep product and regulatory timelines on track. Simulations that ran for days slowed everything downstream and, in the end, limited the number of patients NorthStar could help.

The team’s older on-prem computers didn’t have the space or power for newer, faster hardware. IT had limited HPC expertise, so the team couldn’t take full advantage of MCNP’s parallel computing capabilities. Public cloud wasn’t a clean answer either, since it brought added cost and compliance concerns.

Four points of friction stood out:

1. No cost-predictable way to scale. Expanding on-prem meant new power and cooling. Public cloud meant unpredictable pricing for ongoing simulations.
2. Slow simulation results. Unoptimized computing resources left MCNP jobs running for days and delayed research insights.
3. Throughput limited by capacity and tuning. Aging clusters and workstations, without HPC optimizations like tuned compilers and libraries, MPI/OpenMP, and fast I/O, created frequent bottlenecks.
4. An expertise and compliance gap. The team didn’t have HPC specialists to stand up a compliant, optimized MCNP environment, and manual administration added complexity.

The business impact was real. Without same-day answers, NorthStar faced delayed results, fewer design points per week, and rising costs, all of which put pressure on product readiness and customer commitments.

The solution: a turn-key private cloud built for MCNP

Hoonify delivered TurbOS® as a turn-key private cloud optimized for MCNP. It combined parallel performance tuning, managed operations, and predictable capacity, so NorthStar got same-day simulation results without adding staff, overhead, or hardware. NorthStar engineers now run their own complex simulations, and the company no longer pays for extra subscriptions and outside technical help.

How it works

MCNP simulation input. NorthStar keeps using its existing MCNP models and inputs, including geometry, materials, physics cards, and particle histories. Files are uploaded securely, application access is authorized, and the user workflow doesn’t change.

TurbOS parallel processing and scheduling. TurbOS maps MCNP’s parallel model onto the platform for immediate efficiency. It handles resource allocation, job scheduling, and workflow management, with optimized compilers, libraries, and memory placement backed by repeatable runbooks. A simple dashboard lets researchers run jobs interactively without an IT admin.

Private cloud. Dedicated capacity is sized for MCNP throughput, with no new on-prem racks. High-density compute nodes run on fast scratch and parallel storage inside an isolated environment with enhanced security controls and auditability. The cost model is predictable and scales as needed.

Simulation output. Same models, dramatically faster results. Simulations run 96× faster (4 days to 1 hour) with on-demand access, at 50% lower cost than standard public cloud, and milestones stay on schedule with no added staff or compute resources.

“It ran so fast I thought it was broken.”Josh Peterson-Droogh, Nuclear Engineering & Physics Manager, NorthStar

Implementation: live in one day

NorthStar moved from older on-prem computing resources to an optimized private cloud in a single day. The transition ran in five steps.

1. Scope and planning. Confirmed timelines, simulation workloads, success criteria, and compliance requirements, then finalized the transition plan, partnership roles, and day-by-day schedule.
2. Access setup. Provisioned VPN accounts and secure access, configured firewalls and data paths, and enabled cyber security auditing so researchers could submit MCNP jobs securely on day one.
3. TurbOS deployment. Installed and configured TurbOS on private cloud resources, including SLURM for efficient resource use, MCNP modules and containers, optimized storage, and automation tools.
4. MCNP license integration. Validated authorized software licenses, enforced license-aware access controls, completed RSICC compliance steps, and documented the compliant workflow for ongoing operations.
5. Training and optimization. Validated the performance gains, then delivered hands-on training, job tuning, and job templates before go-live.

The results: a near-100× improvement in MCNP simulation performance

MCNP simulations that took four days now finish in under an hour, a 96× improvement. Research throughput is up 33%, and costs are 50% lower than standard public cloud. NorthStar engineers now spend 80% of their time on simulation work, up from 60%, while IT involvement dropped 20%.

Under the hood, NorthStar’s MCNP input files go through TurbOS for resource allocation, job scheduling, parallel processing, and compliance integration. Jobs then run on Intel Xeon-based HPC nodes with dense compute capacity, no thermal throttling, and predictable scaling costs. Faster turnaround means faster research and earlier clinical trials.

“Time is our biggest enemy. Having an optimized system that provides accurate results and a 100% improvement in simulation performance is critical. Analysis is the bottleneck in our pipeline, and this improved HPC performance significantly enhances our workflow efficiency, helping NorthStar innovate faster and maintain our competitive edge.”Josh Peterson-Droogh, Nuclear Engineering & Physics Manager, NorthStar

Faster answers mean more patients helped

Analysis was the bottleneck in NorthStar’s pipeline. Removing it speeds up radiopharmaceutical development and brings earlier clinical trials within reach.

Hoonify’s team comes from U.S. Department of Energy national laboratories, where they built secure, sovereign computing for high-fidelity modeling, simulation, and AI/ML analysis. TurbOS makes that kind of HPC accessible to research teams like NorthStar’s, with a consistent, compliant stack and a predictable cost model.

Request a TurbOS® demo to see what it could do for your simulation workloads.