JARVIS - Craft / Logic Inc

JARVIS is a heavily proprietary database-application that leverages market and sentiment data to provide price-based prescriptive analytics within the cryptocurrency markets. Conceptualized in 2016, JARVIS has evolved from an experimental concept, into a consumer-facing MVP, to a full Product Release, and is now an internal private application. JARVIS displays remarkable accuracy in forecasting price movements, across a dozen cryptocurrency exchanges, from ingesting sentiment and press content from dozens of sources.

Skills used to design, delivery, and manage JARVIS include machine learning, natural language processing, quantitative development, data science, business intelligence, and various programming languages including Python, PHP, MySQL, PostgreSQL, and RESTful APIs. JARVIS was also developed with a proprietary “onion” architecture, making it remarkably secure, while maintaining accessibility over the web to those with access.

Mock, Concept Art

Simplified Output, Concept

Iteration, MVP

Sentiment Tracking, POC

Forecast Modeling, Beta

Grafana Integration, Internalized

Services: Laravel, PHP, Python, MySQL, PostgreSQL, Vue.JS, AWS EC2, Multi-regional AWS, RDS/Aurora, WAF, ALB configuration, Penetration Testing, Machine Learning/Algorithm Development and Design, Database Architecture, Brand Strategy, Digital Strategy, User Experience Design, User Interface Design, UX Optimization, Application Security.

2016
Conceptualization
Under project name “NET”, JARVIS is conceptualized in 2016 with the belief that sentiment data, sentiment response, and language within press and user generated content, at scale, is reflected in cryptocurrency markets in a statistically reliable way. This belief is found to be fact, and C/L was engaged to turn the idea into an automated software to prove it.
2017
Data Aggregation
Early stages of “NET” required C/L to tap into various data sources, leveraging open read-access APIs for the top ten cryptocurrency exchanges, including – but not limited to – Coinbase, Kraken and Binance. Utilizing the same methodologies, C/L creates a targeted RESTful API reading library that allows “NET” to ingest specific information related to the interests of the upcoming algorithms. The database architecture is architected by C/L to adjust and scale into new data sources.

2017
2018
MVP
The first iteration of “NET” becomes available online. The application is automatically ingesting content and market data, calculating NLP, and running analysis against the data to correlate the sentiment response to content and how it is reflected in the market. During manual tests of the algorithms designed by C/L, it is estimated that a forecast output would occur 3-4 times per day per asset. In practice, “NET” is finding commonalities in sentiment data that produces approximately 660 forecasts per asset per week.
2019
Machine Learning
Early versions of “NET” provided consistent forecast data from sentiment in the cryptocurrency markets, providing directionality and longer-form outlook. In order to fine-tune the algorithms, C/L includes a means for the “NET” libraries to learn from it’s previous successes and mistakes. Conditions in market sentiment that result in accurately providing difference, timing, and probability are appended to the calculation library.

2019
2020
The Pivot
“NET” routinely provides on-time, on-price forecasts with remarkable (81%+) accuracy, sustained, through all market conditions. The outcomes from “NET” pushes the founders to remove it from public access and find reliable strategies that could leverage this unique data set in the cryptocurrency markets. JARVIS was born.
2021
JARVIS Takes Shape
Application architecture is added to JARVIS that attempts to automate trade execution on various cryptocurrency exchanges, utilizing the sentiment forecast data as a trigger point to take action. Like the forecast engine, the trading engine is purpose-built by C/L, using a secure communication tunnel, Python, and PostgreSQL to plan, execute, monitor, exit, and log exchange activity for both backtesting scenarios and live market action in real-time.

2021
2022
Formalization
After more than 6 years of active development, tuning, and teaching, JARVIS reliably publishes over 1,000 forecast data points per asset per week with alarming 85%+ accuracy for forecasted market prices ranges between 0.05% and 8.00% fluctuations, with up to six-day outlooks. The sentiment forecasting of JARVIS provides enough foundation and reliability that formalized trading strategies are currently in development and being tested.

2016

Conceptualization

2017

Data Aggregation

2017

2018

MVP

2019

Machine Learning

2019

2020

The Pivot

2021

JARVIS Takes Shape

2021

2022

Formalization

Browse

Contact

Who Are We?

Insights