My Courses
-
Fine-Tuning LLMs with Reinforcement Learning
Explore how Reinforcement Learning fine-tunes LLMs. This guide demystifies PPO, RLHF, RLAIF, DPO, and GRPO, explaining mechanisms, benefits, and use cases for aligning AI with human preferences.
-
Building a Screen-Aware AI with ScreenEnv and Tesseract
Learn how to build screen-aware AI using ScreenEnv and Tesseract for dynamic, real-time screen content understanding.
-
The CXO’s Role in Building Agentic AI-Ready Talent
CXOs must lead talent transformation to build Agentic AI-ready teams through upskilling, mentoring, and applied learning.
-
Redefining Governance, Risk and Compliance for AI Systems with GRC 2.0
As AI systems become more autonomous, organizations face new governance and compliance challenges. This article explores modern GRC approaches focused on explainability, traceability, and ethical alignment.
-
Building A Multi-Agent AI Marketing Assistant with AWS
Generate powerful ad copies with AI! Learn to build a Streamlit app using LlamaIndex & Gemini, then deploy it on AWS EC2 with Docker.
-
A Practitioner’s Guide to Agent Communication Protocol (ACP)
IBM’s Agent Communication Protocol (ACP) is an open standard for seamless agent-to-agent communication.
-
Mastering Data Compression with LLMs via LMCompress
LMCompress uses large language models to achieve state of the art, lossless compression across text, image, audio, and video by approximating Solomonoff induction.
-
Mastering Scientific and Algorithmic Discovery with AlphaEvolve
AlphaEvolve by DeepMind evolves and optimizes code using LLMs and evolutionary algorithms, enabling breakthroughs in science and engineering.
-
A Deep Dive into J1’s Innovative Reinforcement Learning
J1 by Meta AI is a reasoning-focused LLM judge trained with synthetic data and verifiable rewards to deliver unbiased, accurate evaluations—without human labels.
-
A Deep Dive into Absolute Zero: Reinforced Self-play Reasoning with Zero Data
Absolute Zero enables language models to teach themselves complex reasoning through self-play—no human-labeled data required. Discover how AZR learns coding and logic tasks using autonomous task creation, verification, and reinforcement.
