lavu1/lavumweemba

Academic and engineering portfolio for low-resource NLP researcher Lavu Mweemba # Lavu Mweemba — Academic & Engineering Portfolio An academic-first personal website presenting Lavu Mweemba's research in low-resource natural language processing, Bemba–English neural machine translation, responsible AI, and dependable software systems. ## Highlights - Research agenda, thesis results and published research - Transparent English-to-Bemba research-model demonstration - Filterable portfolio of research, AI, mobile, education and enterprise projects - Professional experience, membership, service, awards and training - Downloadable academic, developer and systems-administrator CVs - Responsive Tailwind CSS design, accessible navigation, SEO and Person structured data - Search and AI discovery through JSON-LD, `robots.txt`, `sitemap.xml`, `llms.txt` and `llms-full.txt` ## Live website The canonical public address is lavumweemba.com. Production deployment notes are maintained in `deploy/README.md`. ## Local development ```bash composer install cp .env.example .env php artisan key:generate npm install npm run build php artisan serve ``` ## Translation endpoint The public demo intentionally falls back to a small set of verified phrases when a private inference endpoint is not configured. This prevents unverified text from being presented as research-model output. To connect the private Hugging Face deployment, configure these server-side values in `.env`: ```dotenv HF_MODEL_ID=lavumweemba/bemba-translation-offline HF_INFERENCE_ENDPOINT=your-protected-inference-en… HF_TOKEN=your_server_side_token ``` Never expose `HF_TOKEN` in client-side JavaScript or commit it to Git. ## Verification ```bash php artisan test npm run build ```