--- title: Afterdose emoji: 💊 colorFrom: blue colorTo: purple --- # Afterdose Open-source pharmacovigilance. Independent analysis of public vaccine safety data. ## What we do We build reproducible analysis pipelines on publicly available vaccine safety datasets — primarily the CDC's v-safe system (7.1M free-text entries obtained via FOIA) and the VAERS lot-level reporting data. Everything is open-source, auditable, and built to hold up under scrutiny. We publish data, methodology, and code. Conclusions require independent review before publication. ## Datasets | Dataset | Description | |---------|-------------| | [vsafe-freetext-frequency](https://huggingface.co/datasets/afterdose/vsafe-freetext-frequency) | Frequency analysis of 7,096,394 v-safe free-text entries. What 10M vaccine registrants reported in their own words. | | [vaers-covid-lots](https://huggingface.co/datasets/afterdose/vaers-covid-lots) | Lot-level adverse event profiles from VAERS. Temporal and geographic reporting patterns per lot. | ## Code **[github.com/afterdose/analysis](https://github.com/afterdose/analysis)** — all pipelines, ingestion scripts, and methodology. ## Data sources - **V-safe**: CDC smartphone-based vaccine safety monitoring. 10M+ registrants, 151M check-ins. Free-text data released via FOIA (ICAN v. CDC). - **VAERS**: Vaccine Adverse Event Reporting System. Public download at vaers.hhs.gov. ## Principles 1. **Structured data first** — use what the database recorded before guessing 2. **Reproducible** — every result regenerated from public data + published code 3. **No conclusions without review** — signal detection is hypothesis generation, not proof 4. **Open** — CC-BY-4.0 for data, open source for code