TokenClusterIdentify Narrative Rotations Before They Hit Mainstream
ML-powered wallet clustering reveals which smart money archetypes favor which narratives
Built on validated research: 2,343 wallets analyzed, 34K+ transactions, >95% statistical confidence
TokenCluster Intelligence Dashboard (Beta)
Real-time clustering analysis revealing behavioral patterns across 2,343+ validated smart wallets
TokenCluster Intelligence Dashboard
Live (Beta)Real-time clustering analysis revealing behavioral patterns across 2,343+ validated smart wallets
📊 Wallet Clusters
Win Rate: 68% • ROI: +142%
Win Rate: 82% • ROI: +218%
Win Rate: 71% • ROI: +156%
Win Rate: 45% • ROI: +89%
Win Rate: 58% • ROI: +103%
Based on 33 engineered features • HDBSCAN clustering
🔥 Narrative Momentum
Last 30 Days
Last updated: 2 hours ago • Based on cluster accumulation patterns
⚡ Smart Wallet Activity
23 wallets in "AI Rotators" accumulated $RENDER
Avg position: $3,200
"DeFi Diamond Hands" rotating 12% portfolio Meme → DeFi
Affected wallets: 58
Cluster 4 adding $ARB positions
Avg $4.2K each • Conviction: Medium
8 wallets buying $FET simultaneously
Coordinated signal • Win rate: 82%
15 wallets reducing $MATIC exposure
Avg reduction: 34%
2,343
Validated Wallets
34,034
Transactions
1,495
Tokens Classified
>95%
Confidence
Dashboard mockup • Bot/MEV-filtered data • Statistical validation applied
Dashboard based on analysis of 2,343 validated smart wallets (Sept-Oct 2025). Bot/MEV-filtered data with >95% statistical confidence.
Academic Research Meets Market Reality
Built on validated AI research, launching as a production platform
TokenCluster democratizes institutional-grade crypto narrative intelligence by applying machine learning to smart wallet behavior analysis. We identify which trader types favor which narratives — giving retail investors the same insights previously reserved for hedge funds.
The Problem
By the time narratives trend on Twitter or Reddit, institutional traders have already positioned themselves. Retail investors miss the first 48 hours where 80% of gains occur.
Our Solution
ML-driven clustering identifies which wallet archetypes are accumulating specific narratives - before price action confirms the trend. Data-driven early warning system.
Research Foundation
2,343+
Smart Wallets Analyzed
34,034
Transactions Processed
1.7M+
Balance Snapshots
>95%
Statistical Confidence
Current Stage: Private Beta (Research Phase)
Core data pipeline complete. Clustering analysis in progress. Public beta launching Q1 2026.
Data Pipeline: Production
ML Clustering: In Progress
Public API: Q1 2026
What We're Building
Early Detection
Identify narrative rotations 24-48 hours before mainstream adoption, when 80% of gains are still available.
Behavioral Context
Understand which types of traders are moving, not just that wallets are moving. Context matters.
Real-Time Alerts
Automated notifications when high-conviction clusters (e.g., 82% win rate) accumulate specific tokens.
Built on Production-Grade Infrastructure
Academic rigor meets engineering excellence. Every claim backed by validated data.
Data Pipeline
Smart Wallet Identification
Dune Analytics queries identify 25,161 candidate wallets, filtered to 2,343 tier-1 validated wallets.
▸ Volume thresholds ($1K+)
▸ Activity minimums (10+ trades)
▸ Bot/MEV exclusion filters
Feature Engineering
Extract 33 behavioral features across 5 categories from 34K transactions and 1.7M balance snapshots.
▸ Performance: ROI, win rate, Sharpe ratio
▸ Behavioral: Trading patterns, timing
▸ Concentration: Portfolio HHI, Gini
ML Clustering
HDBSCAN density-based clustering reveals behavioral patterns and wallet archetypes.
▸ Hierarchical density estimation
▸ Silhouette score validation
▸ Cluster-narrative affinity mapping
Intelligence Dashboard
Real-time monitoring of cluster activity and narrative momentum signals.
▸ Live rotation detection
▸ API access (Pro tier)
▸ Historical pattern analysis
Technology Stack
Data Sources
Dune Analytics
Smart wallet identification, DEX queries
Alchemy Web3
Balance snapshots, transfer events
CoinGecko
Token metadata, narrative validation
Etherscan
Price data, on-chain verification
ML/Analytics
HDBSCAN
Density-based clustering
scikit-learn
Feature engineering, validation
pandas/NumPy
Data processing pipeline
FIFO accounting
Accurate win rate calculations
Infrastructure
PostgreSQL
Time-series optimized storage
Python 3.9+
Backend processing
Parquet
56% smaller than CSV
8-module validation
Quality assurance framework
TokenCluster vs Traditional Whale Trackers
❌ Traditional Whale Trackers
- •Show individual transactions
- •No behavioral context
- •No clustering or patterns
- •Manual interpretation required
- •No statistical validation
✅ TokenCluster
- ✓Cluster 2,343+ wallets by behavior
- ✓33 engineered features per wallet
- ✓ML-driven pattern detection
- ✓Automated narrative affinity mapping
- ✓>95% statistical confidence
Quality Assurance
Bot Detection
MEV & high-frequency filtering
FIFO Accounting
Accurate performance metrics
8-Module Validation
Comprehensive quality checks
>95% Confidence
Statistical validation
Meet the Founder
Jose Luis combines production-scale engineering from Google with cutting-edge AI research from his Master's in Generative AI at MBIT School. After watching retail traders consistently miss narrative rotations while institutional players profited, he built TokenCluster to democratize smart money intelligence.
Previously, Jose engineered systems at Google serving millions of users and specialized in AI/ML applications including RAG systems and document extraction. TokenCluster represents his academic research transitioning to production — applying advanced AI techniques to solve a real market inefficiency.
Technical Lead
Google • Frankfurt
Master's Student
Generative AI (MBIT School)
Specialization
System Architecture & AI/ML
Location
Frankfurt, Germany
I built TokenCluster because I was tired of seeing retail traders exit liquidity for whales who had already positioned themselves. The data exists to level the playing field — we just needed to apply rigorous ML to reveal the patterns.
The Market Inefficiency We're Solving
Traditional Tools Show WHAT Happened
Token prices, whale transactions, social sentiment — all reactive indicators. By the time these metrics confirm a trend, institutional players have already positioned themselves.
Retail investors consistently arrive late to narrative rotations, missing the first 48 hours where 80% of gains typically occur.
TokenCluster Shows WHAT'S HAPPENING NOW
Our AI clusters smart money behavior patterns and identifies narrative momentum before price action confirms the trend. See which wallet archetypes are accumulating which narratives in real-time.
Data-driven early warning system with statistical validation.
The crypto market moves in narratives. AI tokens pump together. RWA coins follow similar patterns. DeFi plays move as a pack. TokenCluster helps you identify these rotations using behavioral clustering — not just price charts.
How TokenCluster Works
Identify Smart Wallets
2,343 validated wallets filtered from 25K+ candidates using volume thresholds and bot detection
Cluster Behaviors
Machine learning analyzes 33 features to group wallets into behavioral archetypes
Track Narratives
Real-time monitoring reveals which clusters favor which narratives with momentum scores
Early Adopter Access
Simple pricing for the private beta phase
Free Tier
Limited beta access for testing & feedback
- ✓Daily narrative heatmaps
- ✓Top 5 cluster activity signals
- ✓Community access
$0
Limited spots available
Pro Access
Full platform access during beta period
- ✓Real-time cluster alerts
- ✓Full API access
- ✓Historical data (3 months)
- ✓Priority support
$49/month
Beta pricing • First 100 users lock this rate forever
Early adopters help us refine the product. Lock in lifetime beta pricing while we're still in private beta.
Join the Private Beta
Early adopter spots available for Q1 2026 launch
No spam. Beta updates and early access invitations only.