Overview In the first blog post of this series we have seen how we can represent words or tokens using Word2Vec. However, we also want representations that capture the context and meaning of words within different sentences. A word in the context of one sentence might mean something different than the same word in another sentence. How can we capture that? Concepts like RNNs and LSTMs address this by preserving information about previous words in hidden state vectors. However, they have limitations such as vanishing gradient issues and computational inefficiency due to the need to compute all hidden states sequentially for a given sequence. ...
The Agentic Operating Model (AOM)
As AI agents evolve from tools into autonomous actors, the question shifts from “What can the agent do?” to “How do we govern what it does?” If you have spent time designing Cloud Operating Models, aligning DevOps practices, platform teams, and shared-responsibility (RACI) matrices, this challenge will feel familiar. The difference: agentic AI doesn’t just execute pipelines. It makes decisions. And that demands a new governance framework similar to decisions which are written down and governed in corporate environments. ...
James: Building a Lightweight OpenClaw on WhatsApp
GitHub Repository: bauer-jan/whatsapp-agent James is a lightweight AI assistant that runs on your device. You chat with it through a WhatsApp self-chat. It connects via WhatsApp Web using a QR code. James can schedule tasks and access other systems through MCP, including email, the internet, services, and databases. Similar to OpenClaw, James includes structured system files such as SOUL.md, USER.md, and HEARTBEAT.md to create an identity for both the user and the agent. ...
How does an LLM become an Agent? The Agentic Loop
How does an LLM become an Agent? LLM are stateless, they predict the next token based on a sequence of previous tokens. They dont have memory between calls, no ability to execute code, and no way to reach the outside world. So how do Agents autonomously call tools and make decisions? The answer is surprisingly simple: a loop. Your application code sends a request to the LLM, parses the response, and if the model requested a tool call, executes it locally, appends the result to the conversation, and calls the LLM again. The “intelligence” lives in the model; the “agency” lives in the loop. The diagram below illustrates this. ...
NLP Fundamentals - From Text to Numbers
Overview This post covers the core concepts that power modern language models: Natural Language Processing (NLP) Tokenization Word representation Word2Vec Natural Language Processing (NLP) Natural Language Processing is a computer science field focused on manipulating text with computational resources. Basically, it’s about teaching computers to understand and work with human language. At a high level, NLP tasks can be distinguished into three main fields: Classification The model assigns a single label to the input text from a fixed set of categories. It chooses one category that best matches the text. ...
Stock Analysis with LLMs and AWS Bedrock Agents
GitHub Repository: bauer-jan/stock-analysis-with-llm This project aims to develop an automated system for stock analysis and portfolio management, using different data sources such as balance sheet data, technical indicators, news, industry information, and general market sentiment — powered by large language models (LLMs) like Claude 3 and leveraging the AWS Bedrock infrastructure. Key Features Stock Analyst Module The system runs weekly stock analysis on the S&P 500, Nasdaq 100, and EURO STOXX 50, ranking each stock within its respective industry based on balance sheet information, technical indicators, and relevant news. The LLM ranks stocks within their industries, stores its reasoning, and outputs BUY/SELL recommendations for each stock. All results are stored in a database for further analysis and trend tracking. ...
Hello world.
This is my blog. I will write about things I find interesting.