Skip to main content
< All Topics
Print

Tommy Carstensen’s Epstein Files Archive: Search, Flight Logs, Financial Records and Research Tools

Snapshot

Tommy Carstensen’s Epstein Files archive brings document discovery, visual search, travel records, financial research and testimony into one public research resource. It is an independent, noncommercial project, rather than an official government archive. Its strongest practical feature is the ability to move from a lead toward the document that can support, qualify or disprove it.

For EpsteinWiki sleuths, this is a useful companion to epstein-data. Use the tools to locate evidence, then read the source. A search result, automated transcription or network connection needs that final human check.

ItemOverview
WebsiteTommy Carstensen
Epstein research hubEpstein Files Transparency Act Resources
Primary usesDocument search, citation resolution, image discovery, travel reconstruction, financial research and transcript comparison
Source collectionsThe site describes coverage of DOJ datasets, House Oversight productions, court records and FBI material
Review dateOctober 3, 2026. Counts below are figures displayed by the archive, not an independent census.

Who Is Tommy Carstensen?

Carstensen’s professional homepage describes him as a bioinformatician and data scientist working on clinical data engineering, visual analytics and reproducible machine learning pipelines. It lists research publications and projects involving maps, large datasets and archival material.

His Epstein project applies that technical background to a difficult research problem: records arrive in different formats, carry different identifiers and often need to be compared across separate releases. This article profiles his research resource. It does not suggest that Carstensen was an Epstein associate.


Start Here: Document Search and Citation Tools

ToolWhat It OffersBest Research Use
Search the CorpusText search with AND, OR, NOT, quoted phrases, proximity, prefixes and parentheses.Find names, phrases and related concepts without relying on a summary.
Citation ResolverLooks for production stamps including EFTA, EFTA_, EFTA_R1_, SDNY_GM, HOUSE_OVERSIGHT and DOJ-OGR.Trace a court or congressional citation to the document containing it.
Ask the CorpusPlain language answers generated from retrieved document excerpts, with linked sources.Find starting documents when you do not yet know the best search terms.
Key DocumentsAn index of frequently cited records.Orient yourself before exploring a large theme.
Documents by TypeA route into records organized by document category.Locate a particular class of evidence.

Boolean search accepts queries such as Maxwell AND (blackmail OR extortion), "Zorro Ranch", NEAR(massage appointment, 5) and recruit*. Try full names, surnames, aliases and spelling variants separately. Poor scan quality and handwriting can prevent a relevant page from appearing.

Production stamps require particular care. An EFTA library identifier without an underscore is different from an EFTA_ processing stamp. Removing an underscore can redirect a citation to an unrelated record. The resolver searches for the printed stamp and explains when the result is a containing document rather than a standalone page.

Ask the Corpus identifies its answers as language model output. Retrieved excerpts may omit important records, and generated explanations may overstate what a document says. Follow the citations before using an answer in an article.


People, Employees and Organizations

SectionWhat Researchers Can Explore
PeopleProfiles and document references, with filters for categories, nationality, legal roles, photographs, testimony, flight records, contacts and other appearances.
Payroll and EmployeesStaff roles, employment periods and payroll related material.
OrganizationsFoundations, companies, trusts, banks, universities, grant recipients and other institutions, with labels describing the evidence basis.
Case TimelineDated events for placing records in chronological context.
Academic NetworkInstitutions, academics, research programs and funding records.
Legal NetworkAn entry point into lawyers, firms and legal relationships.

Evidence labels help distinguish a documented payment from an affiliation, a subpoena or a mention in testimony. These are different facts. Receiving a grant or appearing in correspondence does not establish knowledge of abuse.

The people directory is useful for finding where a name appears. Document counts do not measure culpability, and a profile can combine several kinds of records. Check what each cited document actually establishes.

The academic section distinguishes funding from institutional affiliation and explains gaps in surviving grant schedules. A blank funding field is not a reliable statement that no money ever changed hands. For New Mexico research, this provides a route into the Santa Fe Institute material without treating institutional proximity as proof of misconduct.


Flights, Passengers, Aircraft and Properties

The flight database displays 5,835 flights, 14,447 passenger records and 294 identified passengers across a stated 1991 to 2019 span. These are the archive’s combined index figures. They should not be described as a count of Epstein’s personal flights or of trafficking events.

Travel ToolWhat It Adds
Flight LogsSearchable dates, routes, aircraft and passenger entries, with maps and source references.
Passenger IndexNamed private flight passengers and crew, with counts and links into the flight database.
Co Passenger NetworkA graph of shared flights and related travel connections.
Flight Sources and CorroborationSource descriptions, corroboration counts, recorded disagreements and downloadable data.
Commercial Airline BookingsAggregate route information from airline booking records, with passenger identities withheld.
Aircraft and OperatorsThe fleet, operators and related aviation records.
Airports and RoutesAirport and route views for geographic research.
Border CrossingsCustoms records that can help check travel dates and locations.
LocationsProperties and places connected with the documentary record.

The source comparison is especially valuable. It distinguishes original handwritten logs, a third party spreadsheet compilation, commercial bookings, customs records and provisional machine transcriptions. A source count tells you how the index corroborates a record; it does not guarantee that every passenger identification is correct.

The flight log discussion also explains why initials and full names need checking against the scan. A compiler’s expansion of initials is an interpretation. Multiple versions may redact different passengers, and a name missing from one copy may remain visible in another.

For Zorro Ranch research, compare Santa Fe and Albuquerque routes with border records, correspondence and property references. An airport arrival can help establish a travel opportunity. It does not, by itself, establish a ranch visit or an event inside a building.

Commercial booking maps add a separate layer of travel evidence. A booking is not automatically proof that a traveller completed the trip. The aggregate presentation also protects people whose itineraries could expose survivor identities.


Money, Banks, Companies and the Estate

Financial ToolResearch Purpose
Money HubThe starting point for financial and estate related research.
Money NetworkA searchable financial directory with documented connections, flow views and data exports.
Documented AmountsFigures extracted from documents, with currencies, surrounding context and source references.
BanksInstitution specific routes into banking records.
JPMorgan ChaseA dedicated collection of banking and litigation material.
Deutsche BankA dedicated collection concerning its accounts and handling of Epstein related business.
Apollo Global ManagementMaterial organized around Apollo, Leon Black and Epstein.
Will and Estate accessThe Money hub links to its estate feature covering wills, trusts, beneficiaries and executors.

The amounts ledger explicitly separates a number appearing on a page from a verified transfer. It displays 90,964 cited amounts, while publishing a narrower table of selected transaction like entries. A large figure can come from a news clipping, another person’s account, a balance or an extraction error. The page itself warns that it does not resolve who paid whom.

Before reporting a payment, identify the payer, payee, date, currency and transaction type in the original record. Separate invoices from completed payments, and distinguish account balances, asset values, pledges and transfers. Repeated copies of one statement are not repeated transfers.


Testimony, Transcripts and Documentary Comparisons

FeatureWhat It Helps You Examine
TestimonyCongressional interviews, civil depositions and other witness records, with available transcript and source links.
Witness AnswersCounts of statements asserting no memory or no knowledge in transcripts with usable speaker attribution.
Documented ContradictionsThe archive’s comparisons between testimony and other documentary evidence.
Congressional OversightCongressional material and records.
What Congress CitesCitations paired with the congressional claims they were offered to support.
Cases and Court DocketRoutes into case information and filings.

Witness answer statistics are measurements of transcript language. A high rate of memory disclaimers does not prove evasion. The feature excludes video derived sessions without usable speaker labels from its witness specific calculations.

The contradictions index provides leads for checking testimony against records. Its classifications are the project’s analysis, not judicial findings of perjury. Read the complete question and answer, then assess whether the document addresses the same date, event and scope.

For transcription research, record whether text came from an official transcript, captions, machine speech recognition or human review. Preserve page references for official transcripts and timestamps for recordings.


Images, Face Search and Reverse Image Search

Visual ToolWhat It Does
Image GalleryDisplays extracted scans and photographs, with dataset, file identifier, face, redaction and DOJ status filters.
Face SearchCompares a supplied face with the archive’s reference set.
Reverse Image SearchSearches for similar imagery, including photographs, logos, seals and letterheads.
Thumbs.db GalleryDisplays small cached previews extracted from two Windows thumbnail database files.

Gallery names marked with a question mark are software assignments rather than human confirmations, according to the site. Even a plausible match needs checking. An image may be a photograph, a scanned exhibit or a picture embedded in unrelated reading material.

Face Search’s privacy explanation says the photograph is processed locally, while a numerical face representation is sent to the server for comparison. That distinction matters: local processing of the image does not mean no data leaves the device. A returned match remains a possible identification.

Reverse image matching offers a different route when text extraction fails. You can search a whole image or select a complete printed mark. It may help locate related letterheads or repeated pictures, but visual similarity does not establish a shared author or provenance.

The thumbnail collection identifies its sources as EFTA01242503 and EFTA01242504. A cached preview is a small derivative image. It does not independently establish who created, viewed or deliberately saved the original.


Video, Audio and Native Office Files

Media FeatureWhat Is Available
Video GalleryThe page displays 3,427 video files, including 1,003 with sound and 177 transcribed, with filters and source links.
Audio GalleryThe page displays 188 audio files, with available transcripts and metadata downloads.
Office FilesA catalogue of spreadsheets, documents, presentations, database files and other native formats.
Handwritten MaterialLinks to birthday and contact books, letters, journals, handwritten documents and flight logs.
Black BookA route into contact book pages and their transcription.
Birthday BookA route into the birthday collection.
DiariesDiary related records and the project’s interpretation of that material.

The video catalogue describes browser compatible copies alongside preserved originals. It identifies Whisper speech recognition as a transcription method. Converted playback is convenient for review; retain the original source reference when documenting evidence.

Audio transcripts make recordings easier to locate and compare. Confirm important passages by listening, especially names, dates, unclear speakers and overlapping speech. A voicemail can document a tip being offered without establishing that the tip was true.


Communication Records and Courier Shipments

Record SetPotential Research Value
Skype Forensic ExtractionCalls, contacts and messages presented from seized device forensic material.
Phone RecordsTelephone and contact records with document references.
FedEx Courier TimelineSearchable shipments, dates, locations and source receipts.

Courier records can help corroborate office operations and relationships between locations. The timeline displays 4,632 shipments and acknowledges imperfect OCR names. A shipping record does not, without further evidence, identify the contents or prove delivery to a particular person.

Keep attempted calls separate from connected calls. A saved contact is different from a conversation. A message about an arrangement is different from confirmation that it happened. These distinctions are essential when reconstructing a timeline.


Findings, Developments and Community Leads

Document Findings groups the project’s analysis by subjects such as finances, trafficking, political connections, legal obstruction, surveillance, science and Epstein’s death. The page says analysis remains ongoing and distinguishes allegations from adjudicated findings. Its severity labels describe the project’s assessment.

Foreign Government Ties organizes the archive’s analysis of international dealings. The feature distinguishes documented contacts from an adjudicated finding of a registration violation. Recruitment Infrastructure organizes records about schools, agencies and venues. Attendance, recruitment at a location and institutional involvement remain separate claims.

The Eugenics findings collection provides a dedicated route into the project’s reproductive and genetics related analysis. Its theme label should not substitute for determining what each document supports. The handwritten collections also include Letters and Journals and Handwritten Documents, useful when ordinary text search misses a record.

Outcomes organizes reported consequences, including departures from positions, investigations and regulatory action. Check the linked reporting or order before attributing a consequence to a particular document release.

News provides another discovery route. Reddit Watch describes automated monitoring of discussion, cited document identifiers and topic momentum. Popularity is not verification. Follow an identifier to the evidence before treating a thread as reportable.


Auditing Changed Files and Understanding Missing Results

The DOJ audit entry point links the project’s deleted, modified and placeholder related features. Its September 27, 2026 verification notice reported that none of nine previously flagged URLs remained confirmed gone. That is a statement about those checks, not a finding that every record has been released.

No Images Produced compares placeholder PDFs with native counterparts. A PDF carrying a placeholder may point to a separate audio, video or office file. Inspect the corresponding native record before describing evidence as absent.

The FAQ and File Index explain additional problems: interior EFTA pages may belong to a parent PDF, and a published listing may not enumerate every file reachable on the server. A failed guessed URL is insufficient evidence of deletion.

For a suspected change, preserve the exact identifier, source URL, retrieval date and both available versions. Document what changed before proposing a reason. New redactions, inaccessible paths and changed content require different descriptions.


Collection Counts, Downloads and the Yahoo Emails

The File Index offers dataset inventories, metadata, URL lists, extracted text and media information. Its tables distinguish files listed in a manifest from records that yielded extracted page text. This is useful for reusable research, provided the provenance and redactions are preserved.

The homepage, gallery and directory pages display different totals. They can represent different units, coverage or update states. Files, pages, extracted images, indexed people and searchable text are not interchangeable. Cite the page, the unit and the review date instead of repeating one number as the size of the whole archive.

The FAQ’s external dataset discussion points to the separate DDoSecrets Yahoo email collection. This is a reference to material outside the official productions discussed here. Its inclusion as an external resource does not establish authentication of every message or show that every email is searchable on Carstensen’s site.

For Yahoo email research, preserve the original message and attachment provenance separately. Cross check content with official records where possible. Do not assign an EFTA number to an email unless the same record can actually be traced to an EFTA release.


Survivor Protection and Responsible Research

The survivor protection policy describes name and personal information redaction, removal of survivor and minor images, and a reporting process. It welcomes third party reports of exposed information. These are the site’s stated safeguards, rather than an independent assurance that every exposure has been removed.

Use these tools to investigate conduct and institutions while protecting survivors. Do not publish private addresses, contact details, identifying images or identities inferred from redaction differences. A publicly accessible source still requires an editorial decision about what should be reproduced.


Step by Step Research Guide

  • Define one question. Choose a person, event, payment, trip or testimony passage rather than a broad suspicion.
  • Search in several ways. Use phrases, aliases and Boolean queries. Repeat important searches in epstein-data.
  • Resolve the identifier. Keep underscores and prefixes intact. Confirm whether the result is a document or an interior page.
  • Read the original. Check the scan, recording or native file against the extracted text.
  • Compare relevant records. Bring together dates, flights, correspondence, payments and testimony without treating one as automatic proof of another.
  • Record the limits. Note uncertain initials, transcription errors, missing coverage and alternative explanations.
  • Build the EpsteinWiki entry. Preserve the evidence ID, page or timestamp and clean source links. Separate the record’s contents from your conclusion.

Continue through EpsteinWiki’s Investigation Desk, Audio and Video Desk and Contributor Desk. Carstensen’s resource can shorten the path to evidence. The publishable conclusion still rests on what the evidence supports.


Questions for Further Investigation

  • Which important searches return different documents in Carstensen’s archive and epstein-data?
  • Which passenger names or dates remain provisional when compared with the handwritten logs?
  • Which financial figures can be resolved into independently documented payments?
  • Which testimony comparisons survive review of the complete exchange and contemporaneous records?
  • Which suspected gaps reflect source coverage, redactions, parent documents or an indexing problem?
  • Which Yahoo emails can be corroborated against an official production without exposing private people?

Sources

The tool names throughout this article link to their respective pages. The following sources provide the principal documentation for the archive and its methods.

Previous Lisa T
Next Archived Fox News Report on David Copperfield’s Musha Cay Purchase Cited in the Epstein Investigation
Table of Contents