Projects

Confidentiality note: The image shown is a representative example based on publicly available geometry. Any CFD data or design details specific to my work at Aspect Biosystems have been excluded to maintain NDA compliance.

Project Overview

The goal was to characterise the internal shear environment of a Corning ProCulture 500mL disposable spinner flask used for aggregate-based cell culture, and to use that information to identify process parameters that produce equivalent shear profiles across multiple scales. This enabled scale-consistent culture of shear-sensitive cellular spheroids as part of a scalable bioprocess strategy at Aspect Biosystems.

The Challenge

Stirred-tank reactors, particularly paddle impellers, produce highly localized regions of elevated shear and complex flow patterns. Capturing these correctly requires careful geometric representation of the impeller and realistic boundary conditions. The challenge was to predict peak shear regions that correlate with biological outcomes (spheroid viability and morphology) while balancing the computational efficiency and biological accuracy. It was also crucial that the simulations not only converged numerically, but produced physically realistic results.

Roles and Responsibilities

This was a fully independent project conducted end-to-end, involving the following:

• Preparation of representative CAD: Identified and modeled the critical features of the impeller and internal vessel, modeling them in SolidWorks to accurately represent the drivers of shear generation and propagation.
• Model simplification: Analysed the physical setup and justified simplifications — treated the system as single-phase (assumed minimal phase mixing) and steady-state for the purpose of mapping shear fields efficiently.
• CFD setup: Built the fluid volume mesh and solver setup in Ansys Fluent. Selected turbulence models and solver settings appropriate for high‑Re, impeller‑driven flow and ran simulations to convergence (residuals k <= 1e-3). Gut checks performed with hand calculations to ensure reaslistic results were generated
• Post-processing: Computed shear stress (dynamic viscosity × strain rate) and wall shear stress contours on planes and surfaces to locate maximum shear regions inside the vessel.
• Scale testing: Modeled and simulated smaller scale systems in SolidWorks/Fluent to test process parameters and obtain similar shear profiles; designed and conducted wet lab experiments to validate simulation results.
• Process documentation: Drafted descriptive protocols for use of Fluent to rapidly simulate stirred‑tank systems, ensuring knowledge transfer at the end of the engagement.

Outcomes

The CFD analysis successfully identified where peak shear occurs in the spinner flask and allowed rapid in‑silico screening of scale‑up parameters. Using the CFD‑derived parameter sets, lab testing produced spheroid products with consistent morphology across multiple scales, supporting a scale‑consistent manufacturing strategy. The computational approach reduced the need for repeated full‑scale experiments and accelerated the process development timeline.

Takeaways and Learnings

• Geometric simplifications matter. Modeling every detail adds complexity with marginal benefits after a certain point; focus on features that influence key parameters to speed iteration.
• Mesh quality is critical. Though it seems obvious, I saw firsthand how mesh convergence improves shear prediction around the impeller itself
• Practical testing remains essential. Simplified models accelerate iteration but must be validated experimentally to refine parameters, particularly when working with biological systems.

Potential improvements

Given the tight timeline, the model relied on significant geometric simplifications and a single‑phase approach. With more time, I would:

• Implement a transient model to understand shear evolution during impeller start‑up and time‑varying behaviour.
• Introduce two‑phase modeling to assess whether phase mixing meaningfully affects the physics of interest.
• Conduct a full mesh‑independence study to quantify discretization error.
• Evaluate alternative turbulence models (e.g., RSM or k‑ω SST variants) to study sensitivity and improve shear predictions.
• Perform basic uncertainty quantification to evaluate robustness of the simulation.

What we were building (the point)

Determine the practical depth of photocuring for GelMA under 405 nm illumination to evaluate whether a light‑cured implant could reliably polymerize in a spinal cord injury microenvironment. The intent was to close the loop between simulation and experiment and quantify where the simulation was misleading.

Why it was hard

Hydrogels scatter and absorb light; PDMS and fixture materials refract and reflect it. These optical interactions mean the simulated irradiance can differ markedly from the irradiance experienced by the gel in an experiment. Small geometric tolerances or stray light paths produce large errors in apparent cure depth, so the experimental rig had to eliminate indirect illumination and isolate the gel’s intrinsic response.

What I did (role and methods)

• Rig design: CAD'd and 3D‑printed PDMS master molds to cast a reproducible cylindrical GelMA column and iterated geometry to remove edge effects.
• Optical control: Built a light‑blocking shroud and axial illumination fixture to ensure only direct 405 nm light reached the gel column.
• Experiment: Exposed gel columns for controlled doses, sectioned samples, and recorded the cured/uncured boundary for depth measurement.
• Model comparison: Compared measured cure depths with the optical simulation and used the discrepancy to identify missing material optical properties (absorption/scattering/refractive index).

Outcome

Measured cure depths were roughly three times smaller than the simulation predicted. That discrepancy prompted immediate revision of the optical model inputs and delivered a practical understanding of the limitations on in‑situ photocuring. The test jig and shroud are reusable and now standard for follow‑up characterisation.

What I learned

• Control of stray light and substrate optics is critical for accurate photopolymerisation experiments.
• Small fixture geometry changes can qualitatively change measured outcomes; rapid CAD iteration is high‑leverage.
• Simulations must be seeded with measured optical constants — otherwise they risk directional but quantitatively incorrect predictions.

What I would do differently

With more time I would measure GelMA and PDMS absorption and scattering coefficients and the refractive index, then re‑run the optical model. I would also add quantitative microscopy or mechanical testing to replace subjective visual sectioning with a precise depth metric.

Project Overview

On UBC Rover’s Science Subteam, we work to quantify the presence of life and habitability of Mars based on biomarker detection in soil samples. With historical methods relying on bulky onboard spectrophotometers and preloaded cuvettes, I designed microfluidic cartridges to perform encapsulated life‑detection assays. These chips are preloaded with lyophilized stains, contain a micromixer to combine the input sample and stain, and end in a viewing window which enables fluorometric analysis — all with less than 3 mL of sample and dye.

Challenges

The microfluidic chip needed to efficiently and repeatably mix samples and reagents, all within a minimal form factor as possible. Material cost was a significant concern, therefore it was essential to reduce the material needed in non-critical portions of the cartridge. On the instrumentation design side, it was critical to ensure the power supply to the fluorometer’s excitation LED was noise-free and stable. With a highly noisy onboard power supply and inconsistent power, this was a challenging exercise in analog circuit and PCB design.

Roles and Responsibilities

• CAD & simulation: Designed and simulated fluid flow within the micromixer to ensure laminar flow and evaluate pressure drops; modeled the chip and support hardware in SolidWorks.

• Prototyping: Developed printing and post‑processing methodology using an in‑house FormLabs Form3B+ SLA printer, iterating geometries to reduce material use while preserving functionality.
• Excitation LED circuit design: Used Altium to design the power and filtration circuit for the fluorometer LED, adding a step‑down regulator and robust decoupling. Bench‑tested the circuit on perfboard before ordering PCBs.

• Integration: Implemented PCB best practices (component placement, ground planes, decoupling, analog routing) to reduce EMI and maintain signal integrity in low‑light photodetection environments.

Outcomes

Work is ongoing, but the excitation LED circuit has been proven on the benchtop and PCBs have been sent for manufacturing. The microfluidic chip design has been extensively validated for SLA 3D printing, with an optimizing printing and post-processing workflow; next steps include sensor integration and hardware assembly.

Takeaways and Learnings

• Design for manufacturability (DFM) is crucial: manufacturing constraints must inform early design choices.
• Early, domain‑specific simulation accelerates iteration and reduces costly prototype cycles.
• Analog design fundamentals (decoupling, layout, grounding) are essential for reliable optical measurement systems.
• Prototyping on perfboard is valuable for validating circuit concepts before committing to PCB fabrication.

Future improvements

In terms of the fluid simulation improvements, I plan to learn and apply residence time analysis and reaction‑coupled models to better predict diffusive mixing and reaction kinetics of dyes with soil samples inside the chip. I'd also like to perform some more in-depth FEA of the chip itself, to understand if there is room to reduce material but retain strength.

Project Overview

This is the main project at UBC’s Mars Rover engineering design team, where I am a subteam lead. Our challenge is to develop a Mars rover capable of traversing complex terrain, carefully manipulating objects, lifting heavy loads, and testing for the presence of extraterrestrial life, all with as much autonomy as possible.

The Challenge

• The rover must be constructed from as few COTS parts as possible
• The rover must be able to traverse complex terrain as autonomously as possible
• The rover must be capable of interacting with panels, switches, levels, and must be capable of carrying some weight from 5-10kgs.
• The rover must have an onboard payload to detect biomarkers of life in the soil

Roles and Responsibilities

• Design of a Rasberry Pi based serial control monitor for the communications antenna, enabling serial commands to drive a stepper motor, turning an antenna to face the rover at all times
  • Involving: DC Stepper motor control, Raspberry Pi, Python Scripting
• Programming of a multitude of analog and digital probes for environmental readouts, including gas sensors, thermocouples, soil probes (NPK, pH, ORP)
  • Involving: Circuit design, Servo motor control, Arduino, Raspberry Pi, Python, C++
• Design of a microfluidic life detection system (See Instrumented Microfluidic Chip Design project for more detail)
  • Including: Solidworks, ANSYS Fluent, Circuit Design, Photodiodes and Optical Design
• Manufacturing of power connectors for the onboard PDB and power distribution system
  • Including: Soldering, XT connectors, BMSs, Electrical connectivity testing

Outcomes

This project is underway at UBC Rover as we make improvements after every year’s competition, but we have seen a great deal of success as a team. We were one of only 38 teams from across the world invited to the University Rover Challenge, with a stellar acceptance score of 92.64.

Mach number contour — shows the supersonic expansion region and peak Mach near the nozzle exit (~M=2.5)

Project Overview

This was a personal challenge to refresh propulsion basics: over one week I designed a nozzle to achieve Mach 2.5 out of a cold‑gas thruster. I started with hand calculations assuming isentropic, 1‑D ideal gas flow, then modelled the geometry in SolidWorks and simulated it in Ansys Fluent to validate the calculations. Simulation results matched the hand calculations closely while highlighting areas for improvement given longer development time.

The Challenges

Key challenges included:

• Propellant selection: I used 800 psi CO2 canisters for prototyping availability. Calculations showed CO2 approaches sublimation in the expansion — a signal to consider alternative propellants for future iterations.
• Balancing design with timelines: Learning and applying the Method of Characteristics (MoC) was desirable but infeasible in one week, so I prioritized a practical converging–diverging nozzle design instead.
• CFD stability: Supersonic, compressible flows required solver tuning, alternative turbulence models, and careful BCs to reach converged, physically realistic solutions.

Roles

As an independent project I handled all work: hand calculations for choked flow and Mach 2.5 exit, SolidWorks nozzle design suitable for SLA printing, and setup & simulation in Ansys Fluent.

Outcomes

Fluent validated the predicted state through critical nozzle sections; the simulations also revealed thermal constraints and potential failure modes tied to the chosen propellant.

Simulation Figures & Notes

Takeaways and Learnings

• Tradeoffs are central to propulsion design — small efficiency gains often carry cost/weight penalties.
• Physical propellant properties (e.g., sublimation) can dominate design choices.
• Axisymmetric modeling is a useful computational shortcut: simulating half the domain saves compute while preserving key flow features.

Future Plans

• Redesign around a propellant that won't sublimate during expansion.
• Tune exit pressure to avoid under/overexpansion and improve Isp.
• Explore MoC-based nozzle design and design a test jig with pressure transducers to measure thrust on a printed model.