Projects & research

Nocturne brings listening, reading, and comprehension tools into one browser extension. Its audio controls support chapter and paragraph navigation, voice selection, playback speed, and synchronized word highlighting. The immersive reader removes surrounding page clutter and lets readers adjust typography, spacing, and themes.

Groq-powered tools explain or define selected text, paraphrase paragraphs, and answer questions about the page. AI features require the reader’s own Groq API key; they are optional alongside the reading controls. Keyboard shortcuts provide quick access to the reader and AI tools.

Reusable word-replacement lists simplify recurring terminology, while site-specific rules automate preferred reading settings. Built on Manifest V3 with an esbuild production build, Nocturne uses local storage for larger rule sets and synchronized storage for preferences. This work applies my research interest in making complex digital material easier to access and understand.

A learning-experience build applying recommendations from my master's research. The analysis of 30 games identified opportunities to explain climate systems through visuals, sound, adaptive stories, and data-driven simulations, alongside barriers caused by jargon and limited representation. Canopy translates those design findings into maps, graphs, cause-and-effect scenes, an integrated glossary, and more inclusive characters and scenarios. Its structure is designed for modest connections.

Canopy complements the reading software on this site: Gloss supports understanding an article, Nocturne supports reading the page, and Canopy provides the climate lesson those tools can support.

Study companion software built around four parts that share real infrastructure rather than four separate apps bolted together.

Focus runs a Pomodoro-style timer — 25 minutes of work, a 5-minute break, a 15-minute break every fourth session — kept alive in the background so the session survives closing the panel, with a system notification when a phase ends. Flashcards turns a text selection into a card with one click, automatically attaching the source page and title so review material and citations build themselves as you read, rather than being a separate note-taking step.

Review runs those cards through a simplified SM-2 spaced-repetition schedule — grade each card Again, Good, or Easy, and the next-due date adjusts, so cards you know well show up less often and cards you don't come back sooner.

And because unfamiliar terms are one of the biggest drags on momentum while studying dense material, Jargon scans the current page against the same glossary used in Gloss and Scope, and lets you turn any flagged term straight into a flashcard.

Nothing is sent anywhere; flashcards, review schedule, and timer state all stay on the device. Loop is the study layer of the same software family: Nocturne makes the page readable, Gloss makes the claims checkable, Loop makes the learning stick.

Where Nocturne helps one person read a page more easily, Scope is built for the person who made the page in the first place — accessibility-audit software organized into four views rather than one flat report. Audit checks for missing alt text, low-contrast text sampled against real WCAG AA thresholds, broken or skipped heading order, unlabeled form fields, and missing captions or transcripts on video and embedded media — and recognizes common LMS platforms (Canvas, Moodle, Google Classroom, Blackboard) to give course pages more relevant context. Each finding comes with a suggested fix rather than just a flag: a filename-derived draft alt-text suggestion, a note to darken text or lighten a background, a corrected heading order.

Content turns the same readability and jargon-detection approach used in Gloss onto course material instead of news articles — a Flesch-Kincaid-style grade, a scan against the same climate-and-edtech glossary, and a structure read that flags dense or 'wall of text' writing.

Simulate previews the current page as it might look under protanopia, deuteranopia, tritanopia, full grayscale, a low-vision blur/contrast filter, or a rough dyslexia-style spacing approximation — real color-matrix filters and CSS, not illustrations, with an honest caveat that the dyslexia preview is an approximation for perspective-taking, not a validated simulation.

History keeps a local, per-site log of past scan results, so a team can tell whether their fixes are actually reducing the issue count over time, not just whether today's snapshot looks better. Scan results stay on the device: a timestamp and an issue count per site, never the page content. Scope is the authoring counterpart to Nocturne in the same software family.

Most credibility tools stop at 'is this a real news site?' Gloss goes further, and does it with four features that work off the same read of the page rather than four separate tools bolted together. It scores any article in real time on currency, relevance, authority, accuracy, and purpose — the CRAAP framework I used to select and evaluate 30 climate-education games in my master's thesis — turning a research methodology into a heuristic reading, built from on-page signals, that a reader can run on any article, not just the ones I analyzed academically.

Alongside the credibility score, the software scans for climate-related numeric claims — a temperature figure, a CO2 concentration, a sea-level projection — and shows a well-established reference figure next to each one for comparison; it's deliberately built to inform, not to auto-verify, since a false sense of automated fact-checking would be worse than no tool at all.

Because unexplained scientific jargon was one of the clearest comprehension barriers my thesis findings identified, any glossary term on the page gets a plain-language definition, with an option to highlight matches directly in the article.

And because a single article is still a single perspective no matter how well-sourced it is, one click opens a search for other outlets currently covering the same story, so a narrowly-framed piece comes with an easy way to check it against wider coverage. The jargon glossary sits alongside Nocturne’s word-replace and AI define tools, so the software family feels like one set even though each runs independently.

The intended audience is anyone trying to evaluate climate or science information without a research background: students doing coursework, journalists fact-checking quickly, or a curious reader who wants a second opinion before sharing an article. It runs on the page you're already reading. Nothing is stored, and nothing is sent anywhere except the one search request if you choose to look up other coverage.

A generic 'you are connected to the internet' check is close to useless for remote learning — the real question is whether a specific connection can handle a live class without freezing or dropping audio, and what to do about it if it can't. Pulse answers both parts. It reads a quick network health signal (round-trip latency, estimated bandwidth, connection type) and translates the result into something concrete: a plain readiness read for Zoom, Google Classroom, and Microsoft Teams, with video platforms held to a stricter bar than a mostly-static page.

When the answer is marginal or poor, one click saves the page you're on — title and readable text — for offline reading later, so material is still reachable without a connection.

It also keeps a lightweight local log of every check, and once there's a handful of them at different times of day, it surfaces which hours tend to be best and worst, so a student or teacher can plan around the pattern instead of discovering the problem mid-class.

The idea traces directly back to my bachelor's thesis on e-learning infrastructure during COVID-19 in Bangladesh, where the digital divide and rural broadband access weren't abstract policy concerns — they were the reason entire classes couldn't attend school. Pulse is that research turned into a small, practical set of tools: diagnosis, offline access, and pattern awareness working together, reflecting my IT-support background applied to the same access problems the research uncovered.

Saved pages and connection history stay on the device. Pulse is the access layer of the software family that Nocturne, Gloss, Scope, and Loop continue.

An independent research project examining how AI, immersive technologies, and hybrid video are changing teaching and learning. Drawing on the critical framing of UBC's ETEC 511 course, I ask who designs these tools, how they are used, and what they mean for access, educator support, and institutional decision-making.

I compare the three technology areas through adoption, evidence on learning, implementation demands, and equity. UNESCO's 2023 survey of more than 450 schools and universities found that fewer than 10% had formal guidance on generative AI. The OECD's 2026 Digital Education Outlook makes a related distinction: better task performance with AI does not necessarily mean lasting learning, and educational design matters.

The synthesis considers AI as a layer connecting immersive and collaborative tools, alongside questions about accessibility, governance, and teacher preparation. This extends my earlier work on representation and accessibility in climate-education platforms, with practical implications for evaluating new learning technologies.

Full report available on request.

Master's thesis at the University of Calgary (Werklund School of Education), examining the role of digital games in climate change education (CCE). Working within a constructivist framework, I used multimodal discourse analysis to read 30 climate-education games across their textual, visual, audio, and interactive layers, selecting each title with the CRAAP framework (currency, relevance, authority, accuracy, purpose).

The research answered three questions: how games represent climate discourse through their design; how well they handle inclusion and accessibility, including gender representation and technological access; and what the findings mean in practice for educators, developers, and learners.

Findings showed that visual devices — graphs, maps, a forest shrinking as emissions rise — make abstract climate science tangible, and ambient audio (wind, melting ice) deepens emotional engagement, while unexplained scientific jargon and repetitive sound design undercut both for younger players. The analysis identified stronger opportunities for explaining climate systems in adaptive stories and data-driven simulations than in simple point-and-badge reward systems.

The clearest gap: most games use neutral, generic characters and never address the gendered dimensions of climate change — who is most affected by resource scarcity, and who leads on solutions — despite women being more likely to take real-world climate action. The thesis closes with concrete recommendations for game developers (build inclusive characters, simplify without flattening the science), educators (use games to supplement, not replace, curricula), and policymakers, and was presented at three conferences in 2025.

Manuscript available on request.

The equity-focused strand of the master's research: a close read of inclusion and accessibility across the same 30-game corpus, built around research question 2 of the thesis — how are gender dynamics and technological accessibility represented in digital climate games? The pattern was consistent: most games default to neutral or generic player characters, and none directly address the gendered impacts of climate change — the disproportionate effect of resource scarcity on women in vulnerable regions, or women's outsized role in sustainability leadership — even though a few titles gesture at diversity through surface-level character design.

On the accessibility side, unexplained scientific jargon and static or repetitive audio consistently created barriers, particularly for younger and less experienced players, even in games that were otherwise well designed.

The output isn't only critique: the analysis translates into specific recommendations for developers — build gender-aware narratives and characters, and simplify scientific language without losing depth — aimed at closing the gap between who plays these games and who they're actually built for. Those recommendations feed Canopy, Nocturne, and Gloss: the learning experience and software that followed this research.

The methodology-and-evaluation strand of the master's research: selecting and heuristically reading the 30-game corpus that the wider thesis analyzed. Games were screened with the CRAAP framework (currency, relevance, authority, accuracy, purpose) and then read across four layers — text, visuals, audio, and interactivity — to see what each one actually teaches and how clearly. The corpus ranged from strategy titles like Super Energy Apocalypse: Recycled (energy and pollution management) and Sustainable Shaun (resource balancing), to satire like Climate Change Simulator 2018, to data-grounded tools like the UK government's MacKay Carbon Calculator, which simulates climate projections to 2100 for roughly 30 nations, and platforms such as En-ROADS and STOP Disasters! that this site's climate-games directory also features.

Each game was logged for engagement type (serious vs. entertainment), learning focus, and interactivity level, surfacing which design patterns — narrative-driven simulation over static badge systems — could support engagement and understanding; the analysis evaluated design rather than measuring learners' outcomes. That same CRAAP read later became Gloss: scoring currency, relevance, authority, accuracy, and purpose on any article, not only the games in the study.

Graduate research at the University of Calgary asking whether augmented and virtual reality are classroom-ready for environmental and climate learning, grown out of the same literature review that underpins the master's thesis. AR shows real promise for teaching multimodal literacies and making abstract climate concepts concrete, but only with dedicated teacher training behind it. VR research on brief immersive lessons in Italian secondary schools found genuine post-pandemic gains in student interest and engagement, alongside open concerns about cost, access, and classroom readiness. Related work on AR/VR in sustainable-horticulture education pointed the same direction.

Looking forward, the review outlines where VR could go next in CCE — simulating rising sea levels or disappearing forests to make consequences experiential rather than abstract — while staying grounded in a practical conclusion: immersive tools help when the pedagogy and teacher support exist to back them, and a simpler digital tool is often still the better design choice today. That conclusion is the bridge to New Foundations of Educational Technology, which picks up immersive tools again in 2026 — this time as part of a wider convergence with AI and hybrid video.

Bachelor’s thesis at the University of Dhaka: a mixed-methods study of e-learning during COVID-19 in Bangladesh. I surveyed students and educators on the digital divide, the cost of connectivity, and the psychological load of isolated learning. The work mapped the public and private platforms that kept teaching going, and argued that hybrid models would stay after the emergency ended. Pulse, the classroom connectivity software, is the direct practical follow-on: diagnosis and offline access for the same infrastructure problems this thesis named.

Manuscript available on request.