
**phantom-bridge** turns the open web's scattered, real-time news about a public company into a clean, sentiment-scored signal feed. Type a company name; it pulls fresh coverage across seven dimensions โ financials, legal/regulatory, leadership, operations & supply chain, deals, products, and reputation โ ties every story back to that company, scores each one's financial sentiment with a **local** model, and surfaces it in a live dashboard with scheduled email digests for your watchlist. `Python` ยท `Streamlit` ยท `Bright Data Discover` ยท `distilRoBERTa` ยท `SQLite` ยท `Docker` ## What it does - ๐ **One input, holistic coverage** โ a company-anchored prompt library fans out across 7 news categories so you see the whole picture, not just one beat. - ๐ฏ **On-entity, not noise** โ queries lead with the company and results are keyword-filtered to it, so unrelated events don't leak in. - ๐ง **Finance-tuned sentiment, fully local** โ every story scored by a distilRoBERTa model running offline (no LLM, no API cost, no data leaving the box). - ๐ **Live dashboard** โ analyze any company on demand, manage a watchlist, and configure everything in-app. - ๐ฌ **Scheduled email digests** โ a background daemon scrapes, scores, and emails a per-company report on your chosen interval. - ๐ณ **One-command demo** โ `docker compose up` brings up the dashboard and the watcher together.
31 May 2026

Aiming to streamline the creation of 3D mesh (.obj) files from a given picture - we first use a object detection model (yolo-v8) to extract each bounding box(region of interests). Then we remove background from each bounding box by using rembg library . Lastly we create a 3d model from the images produced using the shap-e library . This project can be used by artists / game developers to easily recreate 3d mesh of any object they have seen in a picture . Currently , command line app has been created . I plan to deploy a flask app with a Vue.js frontend in the future . Libraries used in the project - shap-e , ultralytics (yolo-v8) and rembg (background removal)
14 Aug 2023