Investment Pools

A fully autonomous algorithmic trading fund powered by AI, focusing on short-term opportunities in the top 100 digital assets and funding rate arbitrage opportunities between perp-dex and perp-cexes. The fund uses LSTM models, whale wallet tracking, sentimental market, community and news data, reinforcement learning for dynamic signal generation, execution, and portfolio management.


Key Details

Category

Top100, Funding Rate Arbitrage, Yield Optimisation

Vault Structure

Tokenized Vault (ERC-4626)

Deposits

Open at any time

Withdrawals

7-day lock-up period

Strategy

Long/Short, AI-driven signal processing

Investment Horizon

Short Term

Risk Profile

Medium to High

Position Sizing

Max 10% per asset

Performance Fee

20% success fee on net profit

Management Fee

0% for the first year

Security

MCP, Multi-sig. Security audit by Hashlock


AI & Automation

AI Component

Description

Signal Models

LSTM Neural Networks (price, volume)

Sentiment Analysis

Whale wallet tracking, crypto news integration

Risk Management

Reinforcement learning for drawdown & volatility control

Execution

Automated, latency-optimized CEX and DEX trading algorithms

We specialize in:

  • On-chain and off-chain data fusion, including sentiment, price action, and blockchain analytics.

  • Market-neutral strategies leveraging funding rate arbitrage

With a core emphasis on risk-adjusted performance, capital efficiency, and cutting-edge infrastructure, Knidos is positioned to deliver sustainable returns in the rapidly evolving crypto ecosystem.

Our AI-driven signal engine leverages LSTM (Long Short-Term Memory) models trained on real-time volume and price data to detect momentum shifts and hidden patterns in market microstructure. We enhance these signals by tracking whale wallet activity and integrating natural language processing (NLP) to interpret high-impact news across both traditional and blockchain-native channels.

Whale Wallet Network Modeling Using publicly available blockchain data, we will engineer a graph-based network analysis model to track high-value wallet addresses (whales) and analyze their interactions. The objective is to detect patterns and correlations between wallet activities through unsupervised learning techniques such as clustering, graph embeddings (e.g., node2vec), and temporal graph analysis. This network model will serve as the foundation for a machine learning-based predictive engine that anticipates large-scale asset movements before they impact the broader market.

Advanced Crypto Sentiment Analysis This phase focuses on collecting and processing large volumes of unstructured data from social media (Twitter, Reddit), forums, and news sites to perform sentiment analysis tailored specifically for the crypto domain. We will fine-tune transformer-based NLP models on curated datasets to capture subtleties like sarcasm, hype cycles, and coordinated FUD (Fear, Uncertainty, Doubt) campaigns. The aim is to generate a real-time sentiment index that can be integrated with on-chain signals.

AI-Based Trading Bot Development The final phase will integrate outputs from the network model and sentiment engine into a reinforcement learning (RL) based trading agent. This agent will simulate and execute trades in a sandboxed environment before deploying in real-time. The RL model will continuously learn from market feedback, optimizing for metrics such as Sharpe ratio, drawdown control, and latency of response to key signals.

In addition to strategy-level automation, Knidos will develop a fully autonomous portfolio management system that dynamically allocates capital across trading models based on real-time performance, risk exposure, and market conditions. This system will continuously monitors alpha decay, volatility shifts, and correlation changes to rebalance positions, throttle capital, and preserve edge — without human intervention. The result is a self-optimizing, adaptive trading framework that evolves with the market.

Log AI decisions on-chain in hashed formats to allow ecosystem partners to verify model actions without revealing the model itself. Possible partnership with KiteAI or any other L1 focusing on AI model verifications

transaction infra between Avalanche, Hyperliquid, Arbitrum. Integration with Axelar, LayerZero or Wormhole

AI RISK MANAGEMENT

  1. Volatility Regime Detection (Clustering / HMM) DEVELOP unsupervised learning (K-means, DBSCAN, or Hidden Markov Models) to segment the market into volatility regimes:

    1. Low-vol, trending

    2. High-vol, choppy

    3. Mean-reverting, sideways

  2. Stop-Loss Optimization (Reinforcement Learning)

  3. Use RL agents to learn optimal stop-loss thresholds across different strategies and environments.

  4. The agent is rewarded for minimizing drawdown while preserving upside.

Last updated