Open Source Analyst ยท Computational Social Science ยท Hybrid Threats & Organized Crime
I build data pipelines and computational models for the social sciences โ from electoral data to open-source intelligence โ with research-grade standards: reproducibility, audit trails, documentation.
Hybrid threats ยท Organized crime ยท International money laundering ยท Terrorism & armed groups
I work at the intersection of computational social science and security studies:
- Hybrid threats & disinformation โ influence operations, coordinated inauthentic behavior, astroturfing
- Organized crime & illicit networks โ social network analysis of criminal ecosystems, money laundering typologies
- Terrorism & armed groups โ mobilization dynamics, radicalization pathways, network structures
- Transnational communities โ diaspora networks as both risk and resilience factors
I model social phenomena using explicit sociological theory implemented through LLM-powered multi-agent systems โ not black-box prediction. The framework draws on Goffman (identity performance), Weber (rationality regimes), Merton (anomie and adaptation), Bourdieu (field, habitus, capital), and social network theory (Granovetter, Burt โ weak ties, structural holes, brokerage).
Each agent carries a theoretical role (theory-driven behavioral rules) and the simulation is calibrated against observed data and validated with human expertise โ design principles from my research proposal for a multi-agent simulation of influence operations.
- Data engineering for social science โ extraction, cleaning, normalization and analysis of public datasets (Italian electoral history, demographics, EU open data)
- Social Network Analysis โ academic co-authorship, influence networks, community detection (networkx, Gephi, pyvis)
- OSINT & text analysis โ automated pipelines, compliant web collection, content and sentiment analysis
- LLM multi-agent systems โ agents for research, monitoring and analysis, grounded in sociological models
Note: the GitHub stats cards and trophy widgets (external Vercel services) are temporarily suspended โ they will be restored as soon as the services come back online.
| Project | What it does | Status |
|---|---|---|
| TallyHo | Italian electoral time series 1946โtoday, municipality by municipality, from ministerial open data. Python CLI, CSV/JSON/XLSX/Parquet, 90+ tests, CI, Zenodo DOI | โญ main |
| Arachne-Scholar | Local-first knowledge graph engine for academic literature via SVO dependency parsing. Runs fully offline on your own hardware | active |
| hybrid-kg-paper | Experiments on hybrid knowledge graphs (deterministic NLP + LLM) for scientific text. Reproducibility GED 0.0 | research |
| 2023 โ 2026 |
๐ Universitร degli Studi di Torino Laurea Magistrale in Sociologia e Ricerca Sociale โ Sociology Sep 2023 โ Mar 2026 ๐ Grade: 110/110 cum laude ๐ "Economic Networks and Illicit Practices in the Prato District: a Sociological Analysis of the Chinese Presence between Mafia Infiltration and Community Self-Regulation" ๐ ๏ธ Organizational skills ยท Information analysis ยท +3 |
| 2025 |
๐งญ Universitร degli Studi di Torino โ Future Scenarios Lab Apr 2025 โ Jun 2025 ยท Certificate Practical lab building frameworks for future-scenario analysis: identifying which factors/events may correlate and show causal links with a phenomenon in its future development. ๐ ๏ธ Information analysis ยท Risk management ยท +1 |
| 2025 |
๐ค Universitร degli Studi di Torino โ Algorithms & Society Lab Feb 2025 โ Apr 2025 ยท Certificate |
| 2019 โ 2023 |
๐๏ธ Universitร degli Studi di Firenze Scienze Politiche โ Political Science Sep 2019 โ May 2023 ๐ Grade: 103/110 ๐ "New Technologies, Old Inequalities: a Critical Analysis of Digital Human Resources" |
- MicroRack Systems โ founder & operator: data pipelines, research products, multi-agent automation for analysis work
| Language | Level |
|---|---|
| ๐ฎ๐น Italian | Native |
| ๐ฌ๐ง English | B2 |
| ๐ช๐ธ Spanish | B1 |
| ๐จ๐ณ Chinese | A2 (active/passive) |
- GitHub: jack89-ML
- LinkedIn: jacopoperacchio
- Email: jperacchio8@gmail.com
Research-grade standards: reproducible, documented, auditable.