DEV Community: Adam Cameron Drummond

Gemma To-Dos

Adam Cameron Drummond — Thu, 22 Oct 2026 18:25:08 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady

What I Built

Gemma To-Dos is a small web app that stores your to-dos and uses a local Gemma model to make them easier to start. Press Suggest on any item and it returns a few tips for getting it done. It also returns an ordered list of concrete sub-tasks, which you can add to the to-do with one click.

It's for anyone who stares at "Plan the trip" or "Clean the garage" and doesn't know where to begin. The model does the breaking-down, so you only have to do the doing.

Demo

Code

AdamDIOM / HacktoberfestOnlineProjects

How I Built It

I built this software using Copilot.

Model: Google Gemma (gemma3 by default), run locally through Ollama. You can swap it with OLLAMA_MODEL.
Backend: one dependency-free Node file (server.js) using the built-in http module and fetch. To-dos are saved to a local todos.json.
Frontend: a single index.html with vanilla JS. All text is rendered with textContent, so model output can't inject HTML.
Structured output: the call to Ollama's /api/chat passes a JSON schema in format, requiring tips and subtasks arrays. This keeps Gemma's answer parseable, so the UI can show tips and turn sub-tasks into real checkable items instead of scraping free text.
Flow: POST /api/todos/:id/suggest asks Gemma and stores the result on the to-do. The sub-tasks only become real ones when you click "Add all as sub-tasks", so the model never changes your list on its own.

Why Does Open Innovation Matter?

To-dos are personal: "call the doctor", "finish the job application". With an open-weight model running through Ollama, none of that text leaves my laptop. There's also no API key, no per-request billing and no rate limit. It works offline, and I can swap Gemma for another model by changing one environment variable. A closed API would have meant sending my private task list to a third party and paying for every suggestion.

Prize Categories

Best use of Gemma

OpenTriage: Solving IT Helpdesk's Everlasting Issues

Adam Cameron Drummond — Tue, 13 Oct 2026 23:36:27 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle

What I Built

Using Cursor, I built OpenTriage, a request desk for small organisations that need a clearer way to handle internal IT issues.

Employees submit a request with its details and impact. A locally running model classifies it by category and urgency, assigns a score, writes a short summary, and suggests an owning team. The IT inbox can then be filtered by category, urgency, and status, with controls to update a request or run the classification again.

This could be useful to a small business or nonprofit without a dedicated service desk, or to an IT team that wants a lightweight triage tool. The potential customer is the organisation paying to spend less time sorting requests by hand.

Demo

Code

Check out the folder 2 here:

AdamDIOM / HacktoberfestOnlineProjects

How I Built It

OpenTriage uses Llama 3.2 3B through Ollama, running locally. The app is built with FastAPI and Pydantic, with a small JavaScript frontend and SQLite storage.

The classifier sends the request details and a category and urgency rubric to Ollama. It asks for a response matching a JSON Schema, then validates the result against a Pydantic model and retries if the output doesn’t validate. If Ollama is unavailable, the request is still saved and shown as unclassified so it doesn’t disappear from the queue.

The model classifies and summarises; the IT team still reviews requests and manages their status.

Why Does Open Innovation Matter?

IT requests can contain names, work email addresses, and details about internal systems. Running inference through Ollama on the organisation’s own machine means those request details don’t need to be sent to a third-party AI API. The team can also choose a different locally available model through configuration.

That control makes this design practical for organisations that want AI-assisted sorting but aren’t comfortable sending internal request data to an external service. OpenTriage is a small example of how open models and tools can be assembled into a workflow that an organisation can inspect and adapt.

Planespotter.py

Adam Cameron Drummond — Sun, 11 Oct 2026 00:06:02 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass

What I Built (with Copilot)

Planespotter is an offline chatbot that identifies aircraft from a vague description. You type something like "huge plane with a hump on the front and four engines" and it replies with "That's most likely a Boeing 747" and tells you what to check to be sure.

Some of the best places to watch planes, like fence lines, hilltops and the end of a runway, have poor signal or no Wi-Fi. A tool that needs the cloud fails there. This one runs entirely on your own machine, so it works wherever you can stand.

It's aimed at beginners who don't know the vocabulary yet. You don't need to say "high-wing twin turboprop with a T-tail". You can say "two propellers, wings on top", and the bot teaches the words as it goes. The point is to get you looking up at the sky and learning to read planes, not scrolling a screen.

Demo

Example interaction

You: i'm at the airport and i see a large plane with 2 engines on each side and a big hump at the front
Spotter: Sounds like you've got a great start! Based on your description, here are my top picks for the CANDIDATES:

1. **Boeing 747**: This seems like the most likely match, given your description of a large plane with a big hump at the front and 2 engines on each side. The 747 is an iconic wide-body aircraft with a distinctive upper deck, making it a great fit.
2. **Air Force One**: As a special mention, since it fits your description, it's worth noting that Air Force One is a Boeing 747-200B or 747-8, and it often carries 2 engines on each side.
3. **FedEx Boeing 747-8F**: Another possibility is the FedEx Boeing 747-8F, which is a variant of the 747-8 with a cargo-only configuration.

To help narrow it down, take a closer look at the following features:

* Engine count and placement: Are there 2 engines on each side, or are there more engines?
* Wing position: Are the wings positioned on the sides, or are they more offset?
* Size: Is the plane extremely large, or is it proportionally large?

Let me know if you have any more details or if you'd like me to ask a follow-up question!

Code

AdamDIOM / HacktoberfestOnlineProjects

How I Built It

There are two layers, and they work together.

A local aircraft list. About 28 common aircraft (airliners, regional jets, turboprops, private planes, military planes and helicopters) are stored with their engine layout, size, distinctive features and common airlines.
A local open-weight model through Ollama. The default is llama3.2:3b, which is small enough for a laptop. It is configurable through the PLANE_MODEL environment variable.

The flow for each message:

Your description is turned into keywords, with plain-language synonyms mapped to the list's vocabulary ("humpback" to "hump", "chopper" to "helicopter", "twin" to 2 engines).
The list is scored against those keywords, with engine and feature text weighted highest.
The top three matches are given to the model as candidates. It picks the likeliest, explains which features to check, and asks a follow-up question if you're too vague.
Earlier messages in the chat are kept, so a follow-up like "it had a T-tail" narrows things down.

I did it this way because small models are unreliable at recalling aircraft facts from memory. Giving the model a short list of real candidates means it only has to choose and explain, not remember. If Ollama isn't running, the app still prints the matches from the list, so it degrades gracefully.

So far I've checked the matching step on sample descriptions. "Hump and four engines" ranks the 747 first, "two propellers, high wing" ranks the ATR 72 first, and "wings on top with a strut" ranks the Cessna 172 first. Next I want to measure the full chat on a larger set of descriptions.

Why Does Open Innovation Matter?

An offline tool can't call a closed API. Without a connection, a hosted model is simply unavailable. An open-weight model I can download once and run locally made the whole idea possible.

It also does the following:

It's free to use. There are no per-question costs, so a beginner can ask hundreds of silly questions.
It's private and dependable. Nothing leaves the device, and the app can't break because a provider changes or retires a model.
It's changeable. I can swap in a different model, or a smaller one for a phone, with a single setting. Anyone can also add aircraft to the JSON file, such as local airlines and regional types, so the knowledge belongs to the community instead of a vendor. The next step is putting this on a phone, so the whole thing fits in your pocket at the airfield fence.

Prize Categories

Best Use of GitHub Copilot

UK Hackathon Statistics

Adam Cameron Drummond — Sat, 03 Oct 2026 02:03:59 +0000

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend

What I Built

hackreport is a pipeline that reads public pages about past UK hackathons and reports on the student-led events season by season. For each event it records the organiser, whether it was student-led, and where available the size and number of projects.

I built it for a colleague of mine who was interested in seeing statistics of events across the year.
The problem: there is no single place showing which UK hackathons are student-run, who ran them, and how big they were, and organiser knowledge gets lost as students graduate.

Demo

Check out the results here!

Code

Check out the folder 0 Weekend here:

AdamDIOM / HacktoberfestOnlineProjects

Run it with ./run.sh. It sets up the environment, starts Ollama, pulls the model if missing, and writes data/report.md and data/events.csv.

How I Built It

Model: Gemma 3 (4B), run locally through Ollama.
Collect: Python collectors read the Hackathons UK season pages and MLH season pages. They then follow each event's own website, plus its About or Team page.
- Requests respect robots.txt, are rate limited, and are cached.
- If a site is dead, the pipeline falls back to the Wayback Machine copy from within the same season.
- Devpost blocks scripted access, so it reads pages you save from a browser.
Extract: Gemma gets each page and returns JSON matching a schema. It is told to use only what the text states and to leave unknowns as null, with temperature 0 and a verbatim evidence quote for each classification.
Guardrails: a 4B model guesses student-led from a university name, so code checks the evidence quote.
- student_led: true needs society or student wording in the quote.
- false needs company or charity wording.
- Anything else is demoted to unclear.
Seasons: seasons run September to August and are named by the ending year (2024 = Sep 2023 to Aug 2024).
Merge and report: events seen in several sources are merged, with the organiser's own site taking priority. The report is a per-season summary plus a CSV.

Why Does Open Innovation Matter?

Cost and repeatability: an open-weight model running locally means a pipeline that reads hundreds of pages costs nothing per call and can be re-run whenever the sources change. This is great as it means there is no per-token cost for a volunteer or student organiser who may not have access to LLMs.
Constrained output: local inference with a JSON schema, temperature 0 and a fixed model version makes extraction reproducible, so anyone can rerun it and check the results.
Honest limits: because the model is small and open, I had to build the evidence checks. That makes every classification traceable to a quote on a page.

Prize Categories

Best Use of Gemma
Best Use of GitHub Copilot

How the MLH 2022 Season is Gonna Rock!

Adam Cameron Drummond — Thu, 01 Jul 2021 00:24:42 +0000

What are my plans?

One of my big plans for the 2022 season is to run a hackathon! Be it an MLH member event or not, I would love to run an event and coordinate all elements of a hackathon, be it sponsors, prizes, judging or marketing.

What else am I doing?

Honestly, just vibing. Looking to start at University College Isle of Man to study a BSc in Computer Science. Hopefully some form of internship or computing experiences; travel!

Melons vs Pineapples in BLAHAJgang

Adam Cameron Drummond — Thu, 01 Apr 2021 18:05:31 +0000

Where did the Melons Originate?

The melons originated back in January with the conclusion of Local Hack Day: Build. It was an unexpected event however it has led to become part of the biggest meme of Local Hack Day: Share. It came about when Ryan never thanked BLAHAJ. Jack decided it was important to react with BLAHAJ emotes for every single message in Build chat. I took it upon myself to react with melons. From there, it was war. Every message was a battle to see who could get the most reactions.

Where did the Pineapples Originate?

This all came around because of an April Fools Day Prank on Ryan Swift. Michael (from BLAHAJgang) came up with the idea to prank everyone by chaning from MELONsquad to 🍍PINEAPPLEparty🍍. This culminated in what can only be classed as the most iconic Twitch Stream ever. Find it here!

Why did MELONsquad become part of BLAHAJgang?

Before Local Hack Day: Share, the two leaders decided that it would be more beneficial to have one super-guild rather than two smaller guilds. This has clearly proved to be a success as BLAHAJgang is now top of the leaderboard

Where Can I Join the MELONsquad and 🍍PINEAPPLEparty🍍?

Visit us in Discord by going to https://discord.mlh.io and join the BLAHAJgang guild. Once there, ask for the link to our very own Discord server. Top secret! And that's it!

Being an Influencer at Local Hack Day: Share

Adam Cameron Drummond — Wed, 31 Mar 2021 21:46:25 +0000

About LHD: Share

Local Hack Day: Share is an event run by Major League Hacking. It's a week full of learning, challenges and most importantly, sharing. After all, it is called LHD: SHARE.

Why I Enjoy it so Much

I love LHD, or as some say it should be, Global Hack Week, considering it is now a whole week event. I really enjoy attending this event because

I get to meet so many new people, including both coaches and those taking part in the event, like me.
There is SO MUCH to learn & build during the week!
Guilds like BLAHAJgang help with collaboration across the world.

What is There to Learn?

TONNES! There is something for everyone, be it a simple hello world or developing full stack programs. There's stuff in so many different languages that there has to be something for everyone. Even for non-technical people, there was a workshop just about making a dataset look good.

Why Should YOU Take Part in a Local Hack Day?

LHDs are welcoming to ALL. You should join LHD: Share because you can improve yourself. Be it new connections, learning new things or even just to play some mini events, it is totally worth it!

What is BLAHAJgang?

I'm glad you asked. BLAHAJgang is, as said by the leader, 'MLH's meme of the century'. What does this mean for you? There is gonna be a lot of memeable moments. Some of our events include karaoke, ASMR and a virtual pizza party!

To Conclude

You should join Local Hack Day: Share. Come see the fun we have at BLAHAJgang if you want. We also have thanks MELONsquad

My experience at Local Hack Day: Build

Adam Cameron Drummond — Fri, 15 Jan 2021 15:36:10 +0000

How I started

I entered Local Hack Day as my first hackathon not really knowing much about hackathons - or any of the people from MLH. APIs scared me and I had no idea what Node, Flask, JSON or ZEN was. All I had was my passion to learn and meet new people.

What happened over the week

As the week started I was active on all the socials about MLH and watched the majority of the Twitch streams, being active in the comments and on the Discord. Over this time I have met new contacts and continue to recognise the same names in the chat over multiple streams, and have also taken every opportunity to show off my skills or projects using Twitch, StreamYard (in with the presenters) and YouTube.

The Addiction to Chrome Dino

Since the stream from Jacklyn on Day 2, I have made it my mission to get as many of the MLHers to undertake a live on-stream Chrome Dino game attempt. So far I have had Jack, Chris B and Mara attempt it, each getting scores of 583, 3337 and 469 respectively. I plan to continue this for the rest of the event.

How my Love for BLAHAJ has Developed Over the Week

At the start of the week I had never met BLAHAJ or any of the counterparts. I now know and have lots of love for BLAHAJ and really want one of my own... LHD: Build really leaves a lasting impact.

What I am leaving Local Hack Day: Build with

I am leaving this hackathon with a plethora of new skills and connections. Whilst I still may be scared of APIs, I do at least feel like I know a little bit more about them. My main takeaway from LHD: Build is that I now know how to use GitHub! I feel like I have really learned something with this, and that is quite down to Will's stream on Day 3. Furthermore, I sort of know how to use React and Bootstrap applications thanks to Jack's stream on Day 4.