TLDR
After a vision language model extracted 4,286 flights and 392 passenger names from handwritten logs, name deduplication and cross-domain matching identified 13 entities that appear in both flight records and at least one financial or shipping domain (PAPER TRAIL Project, 2026a). These 13 cross-domain profiles include Tom Pritzker, Nicole Junkermann, and Sandy Berger -- a former National Security Advisor.
From Handwritten Logs to Structured Data
Flight logs for Epstein's aircraft were handwritten by pilots in physical logbooks. The pages are difficult to read -- names are abbreviated, handwriting varies, and some entries are partially obscured. Script 16f used a vision language model (Qwen2.5-VL-7B) to extract structured data from 116 unredacted pages, recovering 4,286 individual flight records and 392 distinct passenger names with zero hallucinated names (PAPER TRAIL Project, 2026b).
A separate run on 100 redacted pages produced 1,119 additional flight records, though with 26 extraction errors -- primarily initials rather than full names, a predictable limitation when the source material itself is partially obscured (PAPER TRAIL Project, 2026b). The unredacted set is the primary analytical source.
Name Resolution
The 392 raw names included duplicates, abbreviations, and spelling variants. Script 31 performed name deduplication using exact matching, alias tables, and fuzzy matching via rapidfuzz -- a library that computes string similarity scores to identify likely matches (PAPER TRAIL Project, 2026c). The 392 raw names resolved to approximately 200 distinct individuals.
Each resolved name was then matched against the entity table in the synthesis engine, which contains entities from wire transfers, FedEx shipments, banking records, and institutional analysis. A match means the same individual (or entity with the same name) appears in at least one other analytical domain (PAPER TRAIL Project, 2026a).
The 13 Cross-Domain Profiles
Thirteen entities appear in both flight records and at least one other domain (PAPER TRAIL Project, 2026a). They divide into three categories.
Central figures. Jeffrey Epstein appears across four domains: bank documents, FedEx shipments, flight logs, and wire transfers -- the broadest cross-domain presence of any entity in the corpus (PAPER TRAIL Project, 2026a). His 253 flight events and 28,031 banking events make him the most documented individual by volume.
Operational staff. Larry Visoski (chief pilot, 440 flight events, 21 FedEx events), Larry Morrison (pilot, 293 flight events, 50 FedEx events), Andrea Mitrovich (188 flight events, 2 FedEx events), David Mullen (152 flight events, 2 FedEx events), and Edwina Simmonds (1 flight event, 6 FedEx events) are individuals whose cross-domain presence reflects operational roles -- receiving FedEx packages at addresses associated with Epstein's properties or aviation operations (PAPER TRAIL Project, 2026a).
Named associates. The remaining entities are the analytically significant group. Tom Pritzker appears on 6 flights and in 1 FedEx shipment. Nicole Junkermann appears on 3 flights and in 1 FedEx shipment. Sandy Berger -- Samuel R. Berger, former National Security Advisor under President Clinton -- appears on 3 flights and in 1 FedEx shipment (PAPER TRAIL Project, 2026a). Nick Simmons appears on 1 flight and in 1 FedEx shipment. Eva Anderson appears on 6 flights and in 1 FedEx shipment.
Two additional entries -- "RICK" and a malformed bank record string -- are likely noise from entity resolution artifacts rather than genuine cross-domain profiles (PAPER TRAIL Project, 2026a).
What Cross-Domain Presence Means
Appearing in both flight logs and FedEx records does not establish wrongdoing. A person who flew on Epstein's aircraft and received a FedEx package at a property associated with Epstein may have been a business associate, social acquaintance, or service provider. The cross-domain match establishes co-occurrence across independent data sources -- the kind of convergent evidence that elevates an entity from background noise to a lead worth investigating further.
All 13 profiles are classified as leads with a confidence-adjusted score of approximately 0.514, below the 0.75 threshold required for classification as a finding (PAPER TRAIL Project, 2026a). The adjusted confidence reflects the Chao1 corpus completeness correction: with 36.3% of the estimated entity population missing, any conclusion based on observed co-occurrence must be discounted for the possibility that unobserved entities would change the pattern.
Integration With the Synthesis Engine
The flight domain is now the fifth analytical domain in the synthesis engine, joining wire transfers, FedEx shipments, bank documents, and institutional analysis (PAPER TRAIL Project, 2026d). Adding the flight domain increased the total cross-domain profile count and added the 13 flight-plus-other entities to the leads queue.
The flight domain's primary contribution is co-traveler analysis -- identifying who flew with whom and when. This temporal and relational data complements the financial domains, which show money flows but not physical co-location. When a person appears on both a flight manifest and a wire transfer on proximate dates, the temporal coincidence strengthens both the flight and financial leads.
References
PAPER TRAIL Project. (2026a). Cross-domain entity profiles [Data set]. _exports/synthesis/entity_cross_domain_profiles.csv
PAPER TRAIL Project. (2026b). VLM flight log extraction [Data set]. _exports/flight_logs/vlm_flight_records_unredacted.csv, vlm_flight_records_redacted.csv
PAPER TRAIL Project. (2026c). Flight log name resolution [Data set]. _exports/flight_logs/name_resolution_map.csv, passengers_resolved.csv, dedup_report.csv
PAPER TRAIL Project. (2026d). Cross-domain synthesis engine [Software]. Script 25b, app/scripts/25_cross_domain_synthesis.py
PAPER TRAIL Project. (2026e). Chao1 completeness estimates [Data set]. _exports/validation/chao1_summary.json
This investigation is part of the SubThesis accountability journalism network.