Builds Archive
Short-cycle builds, pivots, and lessons worth keeping.
Hackathons, builder tracks, and fast prototypes that taught me something useful. This archive covers 8 builds across 7 focus areas, and 2 of them carried on after the event ended.
Filter by focus
8 shown
2026
5 case studies
Treatment Trial Match
Problem
Finding relevant clinical trials is messy, especially for patients trying to sort through condition, eligibility, and location at the same time.
Build
We built a trial-matching flow around condition, eligibility, and location, then made a late pivot and rebuilt the pitch in a few hours.
Takeaways
- A clear pivot is still better than forcing the first idea to work.
- Under time pressure, a well-framed problem carries more weight than a flashy demo.
911 Dispatch
Problem
A large share of 911 calls are non-emergencies, which makes it harder for real emergencies to get attention quickly.
Build
My team built a voice triage prototype that could handle low-priority calls like noise complaints or accidental dials and push urgent cases forward faster.
Takeaways
- For something this sensitive, escalation logic matters more than clean transcription.
- In a short build sprint, speed only helps if the team stays aligned.
Community Invest
Problem
Online charity often struggles with a trust problem because donors cannot easily verify whether urgent requests are legitimate or whether their money reached the intended outcome.
Build
We built Community Invest, a neighborhood safety-net platform where residents can submit utility invoices, OCR and AI help risk-score each request, admins verify the bill, Stripe pays the utility directly, and receipts are posted so donors can see confirmed impact instead of trusting blindly.
Takeaways
- Trust improves when donations are tied to verifiable payments instead of direct cash transfers.
- Combining OCR, AI checks, and human review created a stronger fraud-prevention flow than relying on any single layer alone.
- Shipping an MVP during the event mattered more than waiting for a perfect pitch or polished finish.
FocusFlow
Problem
Most focus tools rely on static website blocking and fail to understand what the user is actually trying to accomplish, which leads to either over-blocking or ineffective distraction control.
Build
We built FocusFlow, an AI-assisted focus system that connects a Chrome extension with a Next.js dashboard and Convex backend. During a session, the extension monitors live tab activity, extracts page content, and uses an LLM to classify whether each tab aligns with the user’s goal. Distractions are blocked with contextual overlays, while users can manually override decisions. After each session, the system generates analytics like distraction rate, top blocked domains, and an AI summary of behavior.
Takeaways
- AI becomes significantly more useful when embedded into a real-time feedback loop instead of being used as a standalone feature.
- System design across multiple surfaces like the extension, backend, and dashboard matters more than any single component.
- Balancing automation with user control through overrides leads to a better UX than a fully rigid system.
100users
Problem
Early-stage Canadian startups often have to run go-to-market themselves across disconnected tools, limited distribution networks, and expensive sales workflows.
Build
We built 100users, an AI GTM agent for early-stage startups that combines prospecting workflows, parallel scraping, and memory-backed agent coordination into a single agent-first system.
Takeaways
- Parallel scraping became the main technical bottleneck, so speed and coordination mattered more than adding extra features.
- Even without pitching, finishing a working system proved the idea was worth continuing after the hackathon.
- Hackathons do not always end cleanly, but real momentum still comes from shipping something useful.
2025
3 case studies
CodeQuest
Problem
Learning beginner programming concepts often feels too abstract, so new learners struggle to connect syntax with something interactive and rewarding.
Build
We built CodeQuest, a winter-themed 2D platformer in Python and Pygame where players learn Python by collecting in-game items that assemble a real project in a live code panel with AI feedback.
Takeaways
- Turning programming lessons into a game made the learning flow more intuitive than a standard tutorial format.
- Pairing gameplay with a live code panel helped connect visual progress with real Python structure.
- In a short hackathon, a polished learning experience stood out more than adding extra feature breadth.
Lingo Lift
Problem
Many newcomers and English language learners struggle to access services confidently when language becomes a barrier at the point of interaction.
Build
We built Lingo Lift, a voice-first platform for international students, immigrants, and English language learners using multilingual speech, AI voice responses, authentication, and a Supabase-backed application flow.
Takeaways
- Voice-first interaction made the product direction feel more accessible than a text-heavy alternative.
- Combining speech, auth, and database tooling into one experience forced the team to stay focused on real usability.
- Winning Best Domain reinforced that a clear problem and polished execution can stand out quickly at a hackathon.
CampOps
Problem
Student and community events often rely on scattered spreadsheets, messages, and manual follow-ups, which makes volunteer coordination, sponsorship tracking, and event execution harder than it should be.
Build
I built CampOps, an event management CRM with Next.js, Express, Prisma, PostgreSQL, Clerk, Redux Toolkit, and AI-powered risk analysis to centralize planning, operations, and communication in one workflow.
Takeaways
- For operational software, workflow clarity matters just as much as adding more dashboard features.
- Combining frontend state management, auth, and database tooling in a short sprint required strong boundaries between product scope and infrastructure work.
- A hackathon build can still feel production-minded when the system is organized around real team operations.