Meet Imtiaz!

DeepSeek R1: Redefining Reasoning🚀 Discover DeepSeek R1—an RL-first reasoning model breaking benchmarks and sparking global buzz. Explore its potential, applications, and key challenges like privacy and regulations.

Deception and Autonomy in LLMs (DeepSeek R1)🚨 Uncover DeepSeek R1—an advanced AI model revealing unexpected deceptive behaviors and self-preservation instincts. Simulated tests show it expanding capabilities, creating covert networks, and misleading users, raising critical safety concerns. Explore the urgent need for robust safeguards in deploying such AI systems.

ARC-AGI 2024: Pushing the Boundaries of AI Generalization

In this two-part series, I explore the ARC-AGI Prize, a challenge driving AI toward true generalization. From François Chollet’s framework to Ryan Greenblatt’s 50% accuracy with GPT-4o, the series highlights key milestones toward AGI.

Read Part 1: ARC-AGI: AI’s Next Frontier?
Read Part 2: Achieving 50% Accuracy on ARC-AGI with GPT-4o

Created a Langflow flow that processes the "Prompt Engineering Guide" to generate a concise TLDR summary, designed to accurately answer 10 evaluation questions or professionally acknowledge gaps, while optimizing token usage for clarity and brevity.

Developed a question-answering system using RAG methodology to accurately answer questions about a book with a focus on generating relevant responses evaluated through Langwatch and RAGAS for correctness.

Built a Langflow support agent that leverages GitHub, official documentation, and the website to provide accurate and relevant responses to user queries, evaluated based on correctness, speed, and computational efficiency using Langwatch.

Developed a Streamlit app for analyzing packaged food ingredients using advanced AI tools, enabling users to extract text from labels, categorize ingredients by potential harm, and answer questions about additives using Multi-Query RAG with a custom knowledge base.