Back to Event
Hacktoberfest Hack Day Dublin x The AI Collective Dublin Chapter
Droichead: bridge the gap to the job that's coming
Project name
Droichead: bridge the gap to the job that's comingDemo url
https://droichead.onrender.com/description
Inspiration AI anxiety is real, and in Ireland it falls hardest on people working in a second language. That includes large Polish and Ukrainian communities, and the international staff who keep Dublin's tech and finance sectors running. The advice they get is generic, English-only and often fear-based. Ireland funds excellent upskilling through Springboard+, Skillnet Ireland and SOLAS eCollege, but most people never connect those programmes to a concrete next role. Our belief is simple. A wish to switch jobs is a fantasy. A wish with a deadline is a goal. A goal with a plan is a strategy. Steady action on that strategy is success. Droichead (Irish for "bridge") turns that into a product. What it does A 90-second check-in. A playful, tap-first survey covers role, industry, experience, the skills you enjoy, how you feel about AI, and the hours you can give. You can also dictate your answers or drop in a PDF CV, which is read in the browser and never uploaded. Pulse. You see real news from the last 7 days from Irish and EU sources (RTÉ, Silicon Republic, the Irish Times), each with one line on what it means for you. Live CSO Ireland unemployment figures sit beside a calm read on how your role is shifting. Roles to grow into. Emerging AI-era roles (Forward Deployed Engineer, AI Solutions Engineer, AI-enabled Finance Analyst and more) are ranked by honest fit. Gap analysis. It always leads with what you already have, then what's partly there, then what to build. Bridge the gap. Pick a goal date and get a phased, week-by-week strategy sized to your hours. Every course comes from a human-checked catalogue that prefers free and government-funded Irish programmes, so the AI cannot invent links. My plan. This includes: a timeline with progress, confetti and a weekly streak an .ics calendar export a printable wall planner (goal poster, month calendars, weekly checklist, and QR codes that link straight to each course) live job openings with how long ago each was posted LinkedIn post drafts local events a weekly check-in that re-plans around your real week, without guilt Seven languages: English, Gaeilge, Polski, Українська, Español, Deutsch and Français. Both the interface and every AI answer use the language you choose, and there's read-aloud for accessibility. How we built it Open-weight AI throughout. Every AI call uses Google's open-weight Gemma 4. The hosted demo runs gemma-4-26b-a4b-it through Google AI Studio with thinking set to minimal (a gap analysis takes about 4 seconds). The same code runs fully offline on a laptop with gemma4:12b through Ollama, where your answers never leave the machine. lib/llm.ts also works with any OpenAI-compatible endpoint that serves open-weight models. An Agent Skill. skills/bridge-the-gap/SKILL.md packages the gap-analysis and planning method as an Agent Skill, so any agent can run the same method. Next.js 16, React 19, Tailwind v4 and Motion, with a scroll-drawn bridge story and a floating glass nav. Private by design. There are no accounts, no database and no tracking. Profiles and plans live in the browser's IndexedDB, with one-tap export and delete. The server routes are stateless. Grounded output. RSS feeds supply the news, the CSO PxStat API supplies the statistics, Arbeitnow and Remotive supply live jobs, and courses come from a curated catalogue. The model only ever picks by id. Every response is validated against a zod schema and retried once, with offline fallbacks so a demo never dead-ends. Deployment: Render, using a Blueprint (render.yaml). Challenges we ran into Our first cloud provider fell through over payment setup mid-hack. We moved to a local Gemma 4 on Ollama, then to Gemma 4 on Google AI Studio for the public demo. A local model processes one request at a time. We added sharing of identical in-flight requests, cancelled abandoned generations, and split the roles and news requests so the roles appear first. Gemma on Google's API rejects system prompts and spends its token budget thinking. We folded the system prompt into the first user message and set thinking to minimal. Accomplishments we're proud of A real, local problem (AI career anxiety in a multilingual Ireland) with a calm, specific answer rather than a chatbot. Every link a user sees is real: news, courses, CSO data and job postings. A plan you can put on your wall, scan with your phone, and adjust every week. What we learned Open-weight models are good enough for real, multilingual guidance, including Ukrainian and Polish, as long as you constrain them: schemas, curated data, and no freedom to invent URLs. What's next Native-speaker review of the Gaeilge text. Partnering with Skillnet Ireland, Intreo or community groups that support Ukrainian and Polish workers in Ireland. Irish job boards for the live openings. An offline-first PWA for people with patchy data. AI disclosure (required by MLH rules) At runtime: Google's open-weight Gemma 4 (via Google AI Studio, or Ollama locally) generates role suggestions, gap analyses, plans, news commentary, check-in messages and LinkedIn drafts. While building: we used Claude Code (Anthropic) as an AI coding assistant to help write and refactor code, translations and copy. All product decisions, direction and review were ours. [Video Demo](https://youtu.be/sdD40dQ41WM)Technologies Used
python, javascript, css, react, node.js, typescript, github, google-ai-studio, ollama, next.js, gemma, render, claudeChallenges Submitted
- Best Open-Source AI ProjectSUBMITTED
Team Members
- Muhammad Murtuza Hussain