IResearch
Research
How does light get out of a semiconductor, what changes in a brain with Parkinson's, and why does food stop short of the people who need it?
- Accepted or presented at the named venue
- Submitted
- Ongoing, no submission
Photonics · semiconductor physics · optical modelling
Deep-UV light
- Accepted · poster · IWN 2026
If most of the light an LED makes is polarized the wrong way to leave through the top, where does it go, and can the sides of the chip get some of it back?
- Work
- Interface Polarization Ceiling and Sidewall Recovery of TM Emission in Deep-Ultraviolet AlGaN LED Packages
- Versions
- Interface Polarization Ceiling and Sidewall Recovery of TM Emission in Deep-Ultraviolet AlGaN LED Packages
- Summary
- deep-ultraviolet LEDs can disinfect water and air, but most of the light they generate never makes it out of the chip. part of the reason is polarization: in these materials a large share of the light travels sideways, parallel to the layers, where the top surface reflects it straight back in. i've been modelling how much of that light is lost at each interface in the package, and how much the sidewalls can recover.
- My part
- i worked on the optical modelling, including a polarization-resolved photon-extraction budget that tracks TE and TM light separately through the package.
The longer version
In AlGaN quantum wells with high aluminium content, the ordering of the valence bands changes, and spontaneous emission becomes increasingly transverse-magnetic (TM) polarized, with the electric field along the c-axis. TM emission radiates mostly in the plane of the wells, at large angles to the surface normal.
At a semiconductor–air boundary the escape cone is narrow: with a refractive index around 2.5–2.7 in the deep UV, only rays within roughly 22° of the normal can leave. Light travelling near the plane of the wells is therefore trapped by total internal reflection and is eventually absorbed, for example in the p-type contact layers.
The work treats extraction as a polarization-resolved budget: following transverse-electric (TE) and TM light separately through each interface of the package to see where the ceiling on top-surface extraction comes from, and what fraction of laterally travelling TM light the sidewalls can recover.
Research done while at Unionville High School.
Still open · the model is a budget, not a measurement. how closely it tracks real packaged devices is the obvious next question.
Neuroscience · neurodegenerative & cerebrovascular disease
Neuroscience
- Presented · AD/PD 2026
- Presented · Mayo Stroke Review 2026
- Accepted · poster · CMDSC 2026
What can careful measurement tell us about diseases of the brain, and how do you check it well enough to put your name on it?
- Work
- Three co-authored neuroscience abstracts, three separate studies
- Versions
- AD/PD™ 2026 International Conference on Alzheimer's & Parkinson's Diseases and Related Neurological Disorders
- Mayo Clinic 18th Annual Stroke & Cerebrovascular Disease Review
- Canadian Movement Disorders Society Congress (CMDSC) 2026
- Summary
- i'm a co-author on three neuroscience abstracts. each one is a different study, and each went to a different meeting: AD/PD 2026 in Copenhagen, the Mayo Clinic Stroke & Cerebrovascular Disease Review in Amelia Island, Florida, and CMDSC 2026 in Toronto. i presented the posters.
- My part
- co-author on all three (third author on most of my abstracts), and i present the posters.
- With
- A Johnson & Johnson researcher (JLABS), who reviewed the research and contributed a lot to it.
The longer version
The three meetings cover different ground: Alzheimer's, Parkinson's and related neurodegenerative disorders at AD/PD; stroke and cerebrovascular disease at the Mayo Clinic review; and movement disorders at the Canadian Movement Disorders Society Congress.
Titles, questions and methods for each abstract will be added here once the abstracts are public.
Still open · the honest answer to 'what did you find' belongs to the abstracts themselves. they'll be linked here when they're public.
Food security · nutrition · systems research
Food access
- Accepted · poster · ASPEN 2027
- Accepted · poster · GFS 2027
Food can exist in a city and still be out of reach for the people who need it. Where does access actually break down?
- Work
- Co-authored abstracts on food scarcity
- Versions
- ASPEN 2027 Nutrition Science & Practice Conference
- 6th Global Food Security Conference
- Summary
- this started as a self-directed project about food access close to home and turned into co-authored research on food scarcity. two abstracts were accepted for poster presentation: one at ASPEN's Nutrition Science & Practice Conference, where it was recognised among the top 5% of research abstracts, and one at the 6th Global Food Security Conference.
- My part
- co-author on both abstracts (third author on most of my abstracts), presenting the posters in 2027. the research grew out of platepath, a food-access project i started on my own.
- With
- Researchers at the World Bank Group.
The longer version
Food insecurity is usually less about total supply than about access: price, distance, transport, time, and whether people know a resource exists. That makes it a systems problem, where small frictions in one part of the network decide who is left out.
Details of the methods and results will be added once the abstracts are published in the conference proceedings.
Still open · abstracts describe a problem; they don't fix one. platepath is where i try to do the second part.
Science policy · health data
Digital safeguards
- Presented · Science Summit 2026
As public systems run on more and more data, who decides the standards for sharing it safely?
- Work
- Digital Safeguards; Standardizing Data and Best Practices in an Age of Tech
- Versions
- Digital Safeguards; Standardizing Data and Best Practices in an Age of Tech
- Summary
- i was a speaker at a Science Summit 2026 session, held alongside the UN General Assembly, on data standards and safeguards as technology becomes part of public systems.
- My part
- speaker, listed as a researcher from Unionville High School.
Machine learning
Machine learning
- Ongoing · Apr 2024 – present
What does it look like to contribute to research before you can run the whole project yourself?
- Work
- Machine-learning research with PhD researchers
- Summary
- since 2024 i've worked alongside phd researchers on machine-learning projects: reading and reviewing technical material, and contributing to analysis and discussion.
- My part
- reviewing technical material and contributing to research analysis and discussion.
A research group working with NVIDIA.
AI evaluation · health data
Memory & ADHD
- Ongoing
When an AI assistant remembers things about you, does that memory help or hurt someone with a condition like ADHD?
- Work
- How AI models handle memory for people with medical conditions
- Summary
- with MIT Critical Data, i ran a lot of tests across several AI models to see how what they remember shapes their responses for people with medical conditions such as ADHD.
- My part
- i designed and ran the tests across several AI models and analysed how remembered context changed the responses.
The longer version
Assistants increasingly carry memory across a conversation or between sessions. For a user who has disclosed a medical condition, that memory can make answers more useful, or it can quietly change tone, assumptions and advice in ways the user never asked for.
The work compared several models under controlled prompts, varying what the model had been told earlier, and looked at how the responses changed.
Elsewhere on this site