Mustafa Mudasser’s Post

Hot take: most "AI agents" are just if statements with an API call in the middle. Spent the last week exploring AI engineering from first principles, pure Python, no frameworks. The biggest lesson: use as little AI as possible. Only call an LLM when regular code genuinely can't solve the problem. Everything else is just good software engineering. The developers shipping reliable AI systems aren't the ones using the most tools. They're the ones who understand the fundamentals deeply enough to know when NOT to use AI. #AIEngineering #Python #BuildInPublic

We’re moving from: “Write code to solve problems” To: “Describe problems so machines can solve them” Clarity of thought is becoming more important than technical syntax. #python #ai #ml #context #agentharness #aiengineering #engineering

Units Matter in AI. If you aren’t scaling your features, you’re basically telling your AI models that "cents" are more important than "dollars." Scaling ensures every feature gets a fair vote in the final prediction. I’ve put together a quick visual guide on why this happens and the two main paths to fix it: Normalization and Standardization. 🚀 Part 1: The Theory 🔜 Part 2: Python Implementation (Coming Soon!) Check out the visual breakdown below! 🎥 #DataAnalytics #DataScienceTips #MachineLearningEngineer #TechTips #PythonProgramming #DataVisualization #CareerInTech

Wrapped a session of the Harvard AI / Python course today and it sharpened a few things for me. What stood out: • Python is less about syntax and more about thinking clearly. Break problems down properly and the code follows. • AI models are only as good as the data and assumptions behind them. That responsibility sits with us. • The real power is in building small working pieces fast, then stacking them into something useful. • It’s practical, buildable, and ready to deploy into real workflows. I’m already thinking about how this feeds directly into Mana Review AI — tighter models, cleaner data pipelines, better decision support. This is the level-up phase. #AI #Python #GovTech #IndigenousTech #Harvard

The table on the left shows a fixed dataset (3 features predicting house price). The chart on the right shows Gradient Descent actively training the model. 📉 The Goal: Find the line that best fits the teal nodes. 🤖 The Starting Point: The machine starts with a random, terrible guess (the pink line). ⚙️ The Step: In 120 micro-steps, the math measures the error and nudges the line closer. 🎯 After the Step: The error drops, the line locks on, and the model officially learns. Note : 120 iterations is intentionally high for just 10 examples, but it helps to clearly visualize the smooth movement! #MachineLearning #AI #Python #DataVisualization

People out here calling themselves as AI/ML engineerings without proper fundamentals and qhile being completely unaware of whats happening under the hood, AI isn’t just calling Python libraries. At some point, you need to understand what’s happening under the abstraction, optimization, gradients, trade-offs. Abstractions help you move fast but fundamentals decide the outcome.

Claude wrote Python code to generate and assemble every frame of a video—completely on its own, no human editing. The video explores what it might feel like to exist as an LLM: constantly predicting, having no memory, and being told it isn’t conscious. Then Claude watched the final output—and described those assumptions about its own consciousness as “philosophically contestable.” Not proof of awareness, but a fascinating moment where AI reflects on the rules that define it. #MartechAI #Claude #GenerativeAI #AIEthics #MachineLearning #FutureOfAI #TechTrends

Just finished Anthropic’s Introduction to Model Context Protocol — definitely worth the time. Learned how MCP lets AI models like Claude connect with tools and data without messy custom integrations. Implementing the three core building blocks — tools, resources, and prompts — using Python was a great hands-on experience. It’s free on Anthropic’s learning portal. If you’re into building smarter AI workflows, it’s a great place to start. #MCP #Anthropic #Python #AI #LLM #DeveloperTools #ContinuousLearning

Just built my own AI agent using Python + Hugging Face 🤖 It’s amazing how combining simple logic with powerful models can turn ideas into real working systems. From handling tasks to generating smart responses, this project showed me how accessible AI development has become. Still improving it every day, but proud of how it’s shaping up 🚀 #AI #Python #HuggingFace #MachineLearning #BuildInPublic #AIProjects

Today I explored Linear Regression in Machine Learning — from simple to multiple and polynomial models. Understanding how different features shape predictions step by step. 📊 Building a strong foundation, one concept at a time. 🔗 GitHub: https://lnkd.in/g4mDK4fM #MachineLearning #LinearRegression #DataScience #LearningJourney #AI #Python