Applied AI and computer vision CV

Canadian Nuclear Laboratories - Fuel Research Scientist / Imaging Scientist

Chalk River, Ontario, Canada - October 2021 to present

• Lead applied AI and computer vision workflows for 3D volumetric imaging, including segmentation, instance detection, orientation estimation, quantitative metrology, mesh export, and automated reporting.
• Translate research and inspection requirements into traceable Python workflows, figures, measurements, and reports that support technical decisions.
• Direct the CNL XCT laboratory using a Zeiss Xradia 620 Versa X-ray microscope, manage staff, oversee more than 130 samples annually, and support collaborative XCT and diffraction studies.
• Built an automated 3D instance detection and characterization workflow using Laplacian-of-Gaussian filtering and PCA-based orientation estimation to extract volume, sphericity, surface area, local packing density, and object distribution.
• Reduced analysis time by 80 percent by replacing a manual process with a single-click Python workflow that generates quantitative QA reports and 3D visualizations.
• Designed a three-stage cascaded 3D U-Net for large XCT volumes, with semi-automated labels, connected components, edge detection, positional channels, Tversky loss, and full-resolution inference under GPU memory limits.
• Built synthetic XCT training data from real geometry using simulated projection and CT reconstruction to augment limited datasets and improve model generalization.
• Designed, commissioned, and tested a real-time high-speed camera system for inline particle stream tracking and particle-size estimation during fuel fabrication.
• Support IAEA safeguards inspections and verification activities for nuclear fuel inventories.

University of British Columbia - Research and Teaching Assistant

Kelowna, British Columbia - July 2019 to October 2021

• Applied image analysis to 3D X-ray volumes to quantify polymer infiltration changes linked to more than 100 percent improvement in treatment efficiency and improved mechanical strength.
• Taught undergraduate labs in engineering drawing, hand sketching, and 3D modelling with SolidWorks.

CKF Inc. - Project Engineer

Langley, British Columbia - May 2018 to September 2018

• Completed 40+ engineering projects on an active factory floor, including fabrication drawings, plant layouts, repair support, and process improvement work.

NuclearFlowAI

In progress - Python, FastAPI, React, Next.js, RAG patterns

• Building a no-login public demo for nuclear document workflow automation using seeded public CNSC reactor licensing guidance.
• Designed cited retrieval, deterministic answer modes, saved findings, Markdown summary export, retrieval traces, audit logging, and capped public usage.
• Keeping public upload, OCR, ingestion, index rebuild, and admin controls disabled while the project is in development.

Automated 3D XCT analysis workflows

Applied AI workflow - Python, PyTorch, CUDA, MONAI, reporting

• Built automated analysis workflows for large XCT volumes, including segmentation, instance detection, orientation estimation, metrology, mesh export, and QA reporting.
• Reduced analysis time by 80 percent by replacing a manual workflow with a Python workflow that produces measurements, figures, 3D views, and written QA outputs.
• Built synthetic XCT training data and model-development workflows where experimental labels were limited or expensive to produce.

Task planning and scheduling app

In progress - React, TypeScript, Node, Express, Google Cloud Run

• Built task, calendar, reminder, recurrence, focus-mode, local-storage, and settings flows.
• Integrated Gemini through a server-side API proxy with CORS handling, request validation, rate limits, structured errors, and API key isolation.
• Designed structured prompts and safeguards for task breakdowns, duration estimates, schedule ranking, required JSON responses, consent checks, and fallback behavior.

Robotic bin-picking perception pipeline

Open-source industrial perception project - Python, YOLO, Open3D, PyTorch

• Built a perception pipeline that takes RGB-D images through YOLO visible instance segmentation, point-cloud reconstruction, object-level geometry features, and PyTorch MLP ranking.
• Trained and validated the ranker on 2,016 candidate picks, achieving R2 of 0.9307 and Pearson correlation of 0.9661 against the heuristic target.
• Benchmarked the integrated YOLO-plus-ranker pipeline on 375 scenes with 81.9 percent top-3 agreement against the heuristic reference.

BEV road corridor segmentation

Sensor fusion learning project - Argoverse 2

• Fused camera images, LiDAR sweeps, ego pose, calibration data, and HD maps into a unified representation for road corridor segmentation baselines.
• Completed baseline comparisons, failure-mode characterization, and a technical report describing what worked and what to test next.

Journal articles and book chapters

• M. Poschmann, R. Osmond, P. Shreeves, and A. Prudil, "Inferring the Internal Geometry and Performance of Coated Particle Fuel," Journal of Nuclear Materials, 624, 156476, 2026.
• M. Poschmann, A. Prudil, and R. Osmond, "Fuel Performance Simulations of TRISO Particle Geometries Derived from XCT," Journal of Nuclear Materials, 608, 155714, 2025.
• R. Osmond et al., "An improved plastination method for strengthening bamboo culms, without compromising biodegradability," Scientific Reports, 2023.
• D. Dhir, R. Osmond, K. Golovin, and A. S. Milani, "A high-performance hybrid green composite using plastinated bamboo fillers, with reduced environmental degradation effect," Composite Structures, 2022.
• R. Osmond et al., "A Group Multicriteria Decision Making with ANOVA to Select Optimum Parameters of Drilling Flax Fibre Composites: A Case Study," Composites Part C: Open Access, 2021.
• O. Margoto, R. Osmond, K. Golovin, and A. S. Milani, "Drying treatment of natural fiber fabrics for biocomposites production with vacuum infusion: a plastination case study," book chapter, 2025.
• R. Osmond, K. Golovin, and A. S. Milani, "Review of Chemical Treatments of Natural Fibers: A Novel Plastination Approach," book chapter, 2021.

Conference papers, posters, and panels

• I. Gad-el-Hak et al., "Production of Uniformly Sized Liquid Droplets Using a Vibrating Needle for TRISO Fuel Fabrication," ASME IMECE, 2025.
• A. Prasad et al., "Characterisation of Annular UO2 Pellets Fabricated Using Spark Plasma Sintering and Conventional Sintering," preprint, 2026.
• A. Prasad et al., "Exploring spark plasma sintering as a promising method for TRISO compact fabrication," conference paper, 2025.
• A. Prasad et al., "Fabrication of Graphitic Matrix Fuel Compacts Applicable to High Temperature Gas Cooled Reactors," conference paper, 2024.
• A. Prasad et al., "Progress in Additive Manufacturing of Uranium Dioxide Parts," conference paper, 2024.
• A. Prasad et al., "Fabrication of Homogenised TRISO Analogue Fuel for Reactor Physics Testing," TopFuel 2024, Grenoble, France.
• R. Osmond, "Post-Irradiation Examination of TRISO-Based Fuels Using X-ray Computed Tomography: A Review," 42nd Annual Canadian Nuclear Society Conference, 2023.
• H. K. Namburi et al., "Preliminary Investigation into Micro-Mechanical Characterization of Surrogate TRISO Particles," poster, 2022.
• R. Osmond, D. Dhir, A. S. Milani, and K. Golovin, "Optimizing A Plastination Technique For Preserving Natural Fibre Composites," conference paper, 2021.
• AI in Nuclear Science, panel presentation, October 2025.