The Steeped AI founder manually on-boards client's data with other behavior sources into Steeped AI's NLP, ML and Gen AI tech. Founder then provides distilled outputs and strategy on the client's growth goals.

Currently the only front-end of Steeped AI is this informational site.

• However, the current Vue.js app framework and Bootstrap code templates will be used for the future product.
• Low fidelity mockups of 70% of product UI have been created.

• Neural Network Categorical Model: A neural network tensorflow model is used to categorize various complex concepts in text and metrics for any dataset. It uses embedding, LSTM and dense layers and has been scaled to handle 50+ input signals.
• Unsupervised ML Clustering: A lightweight unsupervised ML cluster setup was made with PCA (shaping), K-means (clustering) and tensorflow's tf_idf (text vectorization) to create a distribution based on variance and cluster that distribution accordingly for any formatted dataset.
• AI System for Tasks: A system of scripts uses OpenAI's ChatGPT API, Google's Gemini API and Meta's Llama 2 to automate complex tasks for Steeped AI. The strengths of each Gen AI are used for tasks including summarization, complex sentiment, nuanced categorization, validation and cluster labels.
• Prototype Data Streams: Steeped AI has two data streams that are being continuously enhanced:
  1. Search Trends: The Google Trends API and other API's like serpapi.com is used to retrieve thousands of search queries and their corresponding metrics to find search insight opportunities for clients.
  2. Business Trends: Company SEC filings, Statistics of U.S. Businesses data, the Yahoo Finance API and WikiData Metadata about thousands of companies are used to provide wide reaching business and product information.
• Auto-Combination Calculations: All readable combinations of dimensions + metrics are calculated through a large SQL pipeline. Dimensions are combined based on their metrics and readability. Most used forms of aggregation metrics are calculated and rounded with all dimension combinations.
• Scaled Wikidata Classification / Knowledge: With any input text data, complex + uniform topics, corresponding extensive metadata and a detailed hierarchy that can connect seemingly different concepts can be found using the WikiData API and other search features.
• ML based annotations: Performance enhanced (faster than Gen AI) ML based annotations are used to quickly identify complex concepts within any text data. SpaCy is used to identify parts of speech and entities like names (scrubbing out public figures and PII), locations and organizations. Pattern is used for sentiment and subjectivity.
• NLP and Vectorization: Scaled NLP processes, such as creating mass ngrams and text normalization, are used to maximize the meaning from any text data. Various ways of vectorizing text are also used depending on processing, including tf_idf, Word2vec and Levenstien.
• Modern Postgres Database System: A Postgresql Database and a corresponding system of pipelines within python that use Apache Beam. This allows for maximum flexibility and scale for any form of data processing within and outside of Steeped AI.

Steeped AI will become a cloud-based, full service platform for data driven professionals that allows them to enter website or owned data and receive decision-ready, editable insights packaged for sharing. Every experience will use proprietary AI and ML fine-tuned for solving complex behavioral research problems unique to each client.

• Highly dynamic and interactive UI: Modern cloud web dev will be used to create an experience that allows for seamless uploading via various methods and easily customizable insight collections that translate into exportable sheet, document, slide or html reports.

• Propriety, Fine-Tuned Gen AI: Steeped AI's proprietary insights data and expertise will be used to fine-tune a preferably open sourced, general-purpose Gen AI model like Mistral, Google's Gemma or Meta's Llama for custom insight tasks. Using custom parameters with LoRa and an abundance of our own private training data will allow us to create a multipurpose insight-based Gen AI model owned exclusively by Steeped AI.
• Neural Network Pipelines: A system of automated and optimized neural networks will be used to sort and cluster data for nuanced insight concepts, and to summarize the predictive factors behind those concepts better than any competitor. Concepts will include "is interesting", "is diverse", "is relevant to client", "is opportunity", and "high growth potential". An automated and vendor-based grading system will be maintained for top accuracy and non-stop improvements.
• Expanded Data Streams: Data Streams will be expanded to many different sources and enhanced for scale including:
  1. Maps / Locational Trends: A maps API like Apple Maps will be used to parse and store company location details, housing patterns and traffic trends for various regions.
  2. News + Industry Blogs: News stream libraries like Gdelt and industry blogs will be parsed for keeping track of relevant trends and insights.
  3. Audience and Purchase Trends: Audience studies like the US Census and purchase streams like credit card reports will be used to track audience and shopping trends.
  4. Product Research: Product research aggregators like Mintel can be used to surface product research findings for most industries or product types.
  5. Expanded Search Trends: The search data stream will be expanded with many more APIs into a vast database of hundreds of millions of searches.
  6. Expanded Business Trends: The business data stream will start to include other business information like earning reports and company blogs.

Steeped AI needs $1.5 million to go to market and maximize on the opportunity to solve the large problems experienced by data driven professionals in this study. Nathan could build 80% of the product himself, but to be competitive he needs a cloud budget to scale and the speed of a dev team.

The money would go to:

• Business Manager: Someone to help with the growth strategy and funding of Steeped AI to ensure smooth operations and success.
• AI Model Fine-Tuning Specialist: An expert in training and fine tuning Gen AI models for specialized tasks.
• Data Scientist: A person to help boost accuracy of data processing, ML and clustering procedures.
• Front-end Web Developer: A developer specializing in well designed front-end experiences.
• Cloud and computing resources: Vast datasets will need storage and ML / Gen AI processes will need computing resources.
• Sustainable Digital Marketing: After achieving a strong organic reach with insight content marketing and a robust CRM system, digital marketing will boost sales efforts.

1. Q2 2024:
   • Attempt all preseed options for initial funding.
   • Secure the first 2-5 clients.
   • Search data stream is finalized.
   • News data stream is started.
   • An automated insights newsletter prototype will be made with Steeped AI tech to share insights to subscribers.
2. Q3 2024: Start to personalize newsletter based on user behavior and corresponding content marketing campaign.
   • Reach $500k in funding.
   • Next 5-10 clients are acquired.
   • News and business data streams finalized.
   • Maps data stream started.
   • Manual on-boarding still needed but Steeped AI is much more automated.
   • A scaled content marketing campaign of consistent related insight interactives is started.
3. Q4 2024:
   • Reach $1.5M in funding.
   • Next 20-30 clients are acquired through direct sales campaigns.
   • Maps data stream is finalized.
   • Audience and shopping data stream started.
   • An online front-end MVP is functional for clients.
   • Over 1,000 content marketing insight interactives are online.
4. Q1 2025:
   • Begin Series A seed funding campaign.
   • Next 200-500 clients are acquired through digital marketing.
   • Audience and shopping data stream finalized.
   • Steeped AI becomes a full self service platform.