Blog Posts

Filtering is the original world model

Deriving the latent state and belief state view through Bayesian filtering. Prediction vs correction as distinct operations, and why the Kalman filter is exact Bayesian filtering under linear-Gaussian assumptions.

Read →

The real problem world models solve

We don't need a model, we need a state. Exploring belief states, sufficient statistics, and why end-to-end model-free RL struggles with partial observability.

Read →

We are going to build a world model

Build log for constructing a world model from first principles. Connecting Bayesian filtering, latent dynamical systems, ELBO, imagination rollouts, and control.

Read →

Notes on My Research Taste

From isolated models to measured systems. How my research approach shifted from optimizing for local novelty to building ideas that survive contact with real system constraints.

Read →

Decoding "Defeating Nondeterminism in LLM Inference" by Thinking Machines

An annotated breakdown of Horace He's deep dive into determinism in language models, with side quests into kernels, atomic adds, and batch invariance.

Read →

Game Theory in Self-Driving: Negotiator

Exploring how game theory principles apply to autonomous vehicle decision-making and negotiation strategies.

Read →

What If a Problem is NP-Hard?

Understanding NP-hard problems and practical approaches to tackle computationally intractable challenges.

Read →

Functions, Variables, and Matrix Completion

Exploring the connections between functions, variables, and matrix completion techniques in mathematical modeling.

Read →

Why Multi-Modal Image Fusion is Not Straightforward

Exploring the complexities and challenges in multi-modal image fusion, discussing why this seemingly straightforward task requires careful consideration and advanced techniques.

Read →

Ever Thought About Tokenizers Much?

A deep dive into the world of tokenizers, exploring their importance in natural language processing and how they shape the way machines understand human language.

Read →

Image Recognition Using PCA

Understanding how Principal Component Analysis (PCA) can be applied to image recognition tasks, exploring dimensionality reduction and feature extraction techniques.

Read →

PageRank Algorithm & Linear Algebra

Exploring the mathematical foundations of the PageRank algorithm through the lens of linear algebra, understanding how it revolutionized web search.

Read →

Read more blogs here:

Substack Medium