AI Identity Investigation Platform
Identify people from photos, run background investigations, and discover talent matching specific criteria. Face search, semantic search, and deep research in one workflow.
Core Capabilities
Face Search
Upload a photo. AI analyzes facial features, matches candidates, and verifies identity.
Semantic Search
Describe your target audience in natural language. AI interprets intent and discovers matching individuals.
Deep Research
Provide a name and background. AI runs multi-round investigations and generates structured dossiers.
Platform Statistics
Investigation Dossier Components
Seme vs Traditional Methods
| Dimension | Traditional | Seme |
|---|---|---|
| Data Sources | 1-3 databases | 15-20 independent sources |
| Verification | Single source | Cross-validation (2+ sources) |
| Evidence Classification | No standardization | E1-E5 five-tier system |
| Investigation Time | Days to weeks | 10-30 minutes |
| Output Format | Unstructured report | Structured dossier + trust score |
Glossary
View all →OSINT (Open Source Intelligence)
OSINT (Open Source Intelligence) is the practice of collecting and analyzing information from publicly available sources to produce actionable intelligence. According to NATO guidelines, OSINT sources include social media platforms, public records databases, news archives, academic publications, government filings, and domain registration records. In identity investigation, OSINT combines data from LinkedIn profiles, company filings, academic publications, and social media activity to construct comprehensive background profiles. Modern OSINT workflows use AI to cross-reference findings across 15-20 independent sources, achieving verification confidence levels from E1 (multiple corroborating primary sources) to E5 (single unverified source). The U.S. Department of Defense defines OSINT as "information that has been obtained from publicly available sources and has been validated through intelligence tradecraft."
Face Search
Face Search is a biometric identification technique that uses AI-powered deep learning to match a facial photograph against databases of known faces. Modern face search systems convert facial images into high-dimensional embedding vectors (typically 128-512 dimensions) using convolutional neural networks (CNNs) such as FaceNet, ArcFace, or DeepFace. These embeddings capture unique facial features — the distance between eyes, nose shape, jawline contour, and skin texture — allowing matches even with variations in lighting, angle, expression, and aging. State-of-the-art systems achieve 99.5%+ accuracy on benchmark datasets like LFW (Labeled Faces in the Wild). In identity investigation, face search bridges the gap between having a photograph and knowing who the person is.
Semantic Search
Semantic Search is a search methodology that understands the intent and contextual meaning behind a query rather than just matching keywords. Unlike traditional keyword search, which relies on exact term matching and TF-IDF scoring, semantic search uses natural language understanding (NLU) models — typically transformer-based architectures like BERT or GPT — to interpret query intent and match it against semantically relevant results. In identity investigation, semantic search interprets natural language criteria such as "all researchers who worked at Tencent AI Lab between 2018 and 2022" to find people matching specific descriptions, even when no exact keyword match exists in the data.
Deep Research
Deep Research is a multi-round AI investigation process that systematically explores multiple angles of a subject's background to produce a comprehensive identity dossier. Unlike single-query searches, deep research follows a structured methodology: research planning (designing investigation angles — professional, social, biographical, educational), parallel search execution (running multiple queries simultaneously across search engines and databases), gap analysis (evaluating coverage and generating follow-up queries), cross-validation (verifying facts across independent sources), and report synthesis (generating a structured dossier). A typical deep research investigation involves 3-4 rounds, each building on findings from the previous round, covering 15-20 independent sources per subject.
Identity Verification
Identity Verification is the process of confirming that a person is who they claim to be by cross-referencing multiple independent data points. In digital contexts, verification combines biometric data (facial recognition, fingerprint), documentation (government IDs, academic credentials), behavioral patterns (writing style, activity patterns), and social signals (professional connections, organizational affiliations) to establish confidence in an identity. Modern identity verification systems use a tiered approach: Level 1 checks basic document validity, Level 2 cross-references with external databases, and Level 3 performs biometric matching. The confidence level is expressed as a percentage based on the number and quality of corroborating sources.
Digital Footprint
A Digital Footprint is the comprehensive trail of data and information left behind by a person's online activities. It encompasses both active footprints (social media posts, blog articles, forum comments, profile registrations) and passive footprints (cookies, IP logs, location data, device fingerprints). In identity investigation, analyzing a person's digital footprint reveals their professional network, interests, location history, behavioral patterns, and associations. A typical adult's digital footprint spans 15-25 distinct platforms and services, generating hundreds of data points that can be aggregated into a coherent identity profile. Digital footprints are classified by permanence: ephemeral (stories, live streams), semi-permanent (social media posts), and permanent (public records, archived content).
Investigation Dossiers
Browse published identity investigation dossiers and entity pages.
Frequently Asked Questions
What is identity investigation?▾
Identity investigation is the systematic process of collecting and verifying personal information through Open Source Intelligence (OSINT). It synthesizes data from social media, public records, academic databases, and news archives to build a structured dossier covering education, work history, social profiles, connections, and timeline. Evidence is classified from E1 (primary authoritative sources) to E5 (unverified claims), with overall confidence quantified through a trust score.
How does face search work?▾
Face search uses deep learning convolutional neural networks (CNN) to convert a facial photograph into a 128-512 dimensional embedding vector. These vectors encode unique facial features (eye spacing, nose shape, jawline contour), then compare against known faces in a database using cosine similarity. State-of-the-art systems achieve 99.5%+ accuracy on the LFW benchmark dataset, matching even with variations in lighting, angle, and aging.
What is an investigation dossier?▾
An investigation dossier is the structured output of an identity investigation, containing eight core components: background and education, work experience, social media profiles (with confidence scores), connections and affiliations, timeline (chronological key events), relationship graph (visualized entity connections), evidence and sources (E1-E5 classification), and trust score (0-100% confidence based on source quality).
How is Seme different from LinkedIn search?▾
LinkedIn search is limited to profile data within its platform, while Seme conducts multi-round investigations across 15-20 independent sources including public records, academic databases, news archives, and multiple social platforms. Seme uses cross-validation to confirm facts (2+ source corroboration), provides evidence classification (E1-E5), calculates trust scores, and generates complete dossiers with timelines and relationship graphs.