sumtyme.ai

COMPANY

A team of researchers and engineers focused on building foundational AI models for the financial markets.

Why Foundation Models for Financial Markets?

Current models relying on statistical AI lack a fundamental market understanding, requiring constant retraining as conditions evolve. Our foundation models provide a generalised understanding of market dynamics, serving as a robust base layer users can build upon instead of starting from scratch.

Core Innovation

A single foundational AI model that predicts market trends generalising to all asset classes and time intervals without requiring market-specific training data.

Training Approach

We simulate synthetic price data using proprietary techniques that replicate realistic market behavior. This approach eliminates overfitting risks to specific markets or time intervals.

Our Theory To Modelling Financial Markets

Our theory is all price movements in financial markets originate in high-frequency data before propagating to lower timeframes. A single model architecture that generalises across multiple timeframes once fine-tuned will provide a holistic understanding of market behaviour.

Key Breakthrough

Our models leverage proprietary reasoning capabilities that allow them to understand, anticipate and navigate market conditions through first-principles understanding rather than learned patterns.

Abstract Reasoning

Just as humans learn multiplication principles rather than memorising every calculation, our models understand underlying market dynamics rather than relying on historical patterns. This mathematical abstraction enables effective generalisation across diverse market conditions, even when facing scenarios absent from their training data.

Testing

We've developed a sandbox environment that's free for anyone to use, allowing users to evaluate the performance of our models across various market scenarios.

How To Use Our Base Models

While our base models deliver strong results in zero-shot scenarios, their predictive power can be significantly enhanced through strategic finetuning across multiple timeframes or by incorporating your proprietary data sources.

Questions?

DEVELOPERS

Fast and easy-to-use APIs

Developer Guides API Reference

PYTHON

# Import necessary libraries
import pandas as pd
import requests
import json

# Define function to create json for API
def time_series_dict(df):
return {
"datetime": df['datetime'].tolist(),
"open": df['open'].tolist(),
"high": df['high'].tolist(),
"low": df['low'].tolist(),
"close": df['close'].tolist()
}

# Import Data
df = pd.read_csv("your_file.csv")[['datetime', 'open', 'high', 'low', 'close']]

# Send POST request
response = requests.post(
"https://www.sumtyme.ai/shared/base_model/ts",
headers={"Content-Type": "application/json"},
data=json.dumps(time_series_dict(df))
)

# Print response
print(f"Status: {response.status_code}")
print(f"Response: {response.text}")

Reasoning Over Recall.