CALI TECHNOLOGIESHow Cali interprets, enriches, and explains crypto activity.
This page provides a high-level view of Cali's data pipeline, enrichment logic, risk and typology methodology, AI assistance patterns, and evidence model. It is written for compliance, investigations, security, and technical reviewers. All descriptions are informational only and do not constitute legal or regulatory advice.
Cali distinguishes between different layers of information so investigators can see what is factual, what is inferred, and what is narrative:
- Raw on-chain data (blocks, transactions, logs, traces).
- Normalized activity records (Activity events) with timestamps, hashes, addresses, and amounts.
- Chain metadata such as block heights, fees, and token contract properties.
- Enrichment-based inferences (e.g., potential entity or service categories tied to an address).
- Risk, typology, and behavior flags based on defined patterns and thresholds.
- AI-assisted summaries and narratives that explain observed behavior in human language.
Cali's interface is designed to keep these layers visually and conceptually separate. Investigators can reference underlying factual data directly when drafting reports, SARs/STRs, or legal materials.
Cali ingests and normalizes blockchain data from multiple supported chains, focusing on an investigation-friendly activity model:
- Raw on-chain events are unified into Activity rows per transaction/event.
- Transfers, internal value moves, contract calls, and logs are standardized across chains.
- Address metadata (labels, ENS), token metadata, and protocol context are applied during enrichment.
- Timestamps are normalized (UTC baseline) to support multi-jurisdictional case work.
Centralized & fiat-linked activity
Cali ingests trade blotters, deposits, withdrawals, conversions, and fiat-linked movements and normalizes them into the same investigation activity model used by Activity, Pivot, and Graph.
Cali maintains a library of behavioral patterns used by AML, fraud, and cyber teams. These are investigative signals, not regulatory determinations:
- Structuring and smurfing indicators based on frequency, amount bands, and timing.
- Layering and obfuscation patterns across hops, chains, and intermediaries.
- Signals tied to mixers, bridges, high-risk services, and negative OSINT association.
- Velocity, burst, and abnormal movement patterns relative to historical behavior.
- Qualitative typology hints (e.g., scam-like behavior) based on contextual patterns and OSINT alignment.
- No typology output is a legal determination or a SAR/STR classification by itself — it is an input to human judgment.
- Cali reconstructs inbound and outbound flow paths for addresses, clusters, or entities of interest.
- Paths may span chains, bridges, mixers, and centralized services where supported.
- Complex DeFi interactions (swaps, LP adds/removals) can be abstracted to higher-level edges for readability.
- Investigators can pivot from graph nodes back into Activity rows for detailed timeline review.
- Graph views are optimized for investigative clarity, not full-chain coverage of all connected activity.
Cali aligns open data sources to addresses, entities, and counterparties where possible:
- Sanctions lists (where applicable and available).
- PEP and exposure signals where supported (informational only).
- Adverse media, complaints, and victim report systems.
- Service categorization (exchange, mixer, merchant, etc.) where inference exists.
- OSINT hits are presented as context; institutions remain responsible for their own screening regimes and determinations.
Cali uses AI to help investigators read, organize, and explain complex activity — never to render decisions:
- AI outputs are non-deterministic and may vary between runs.
- AI is used for narrative drafting, risk rationale explanation, and summarization — not for KYC, sanctions screening, or legal classification.
- Each AI output is tied to underlying factual activity that investigators can inspect directly.
- Investigators are expected to review, edit, and take ownership of any narrative before use in official filings or court.
Different parts of Cali behave in different ways with respect to repeatability:
- Raw on-chain data, normalized activity, and timestamps.
- Static enrichment steps driven by known mappings (e.g., certain contract labels).
- Saved case views, filters, and exported structured data.
- AI narratives, rationales, and suggested typology descriptions.
- Some heuristic scoring and pattern ranking, especially where thresholds are configurable.
- Any free-text explanation or summary is considered advisory and must be reviewed by the human user.
- Consistent, normalized timestamps and chain metadata.
- Repeatable enrichment logic tied to the same underlying data snapshot.
- Clear separation of factual records, heuristics, and AI-generated explanations in the user interface.
- Exports designed so that third parties (e.g., auditors, regulators, courts) can follow the reasoning and verify underlying evidence.
- Designed for investigations involving thousands of transactions and multiple chains.
- Virtualized tables in Activity and related modules for responsive scrolling.
- Progressive graph rendering for dense multi-hop, multi-chain flows.
- Where caching is used, it is aimed at performance and does not replace underlying source-of-truth data checks.
Cali is an investigative and intelligence tool. It is important to understand what the platform does not do:
- Cali is not a KYC system, sanctions screening engine, or regulatory reporting system.
- Cali does not perform identity verification or customer due diligence.
- Cali does not make legal, regulatory, or supervisory determinations; those remain with the institution and its advisors.
- All AI outputs are advisory only and must be reviewed and, where appropriate, edited or discarded by investigators.
Cali welcomes conversations with compliance officers, investigators, regulators, and security teams who need to understand how the platform behaves in detail.