Living Timeline
This page is a living timeline that brings together my long-form writing, project launches, and a few personal milestones in one chronological stream.
By listing articles and project releases side by side, you can trace how ideas move from early notes into published posts or shipped software over time. Each entry is tagged with the domains involved — from LLM systems and reinforcement learning to the occasional life event that influenced the work.
Everything is ordered from newest to oldest, with upcoming experiments included to highlight what's currently in the pipeline.
Future plans are clearly marked so you can see what might be coming next.
Activity Feed
2026
- PPO for Language Models: The RLHF Workhorse (opens project or article)
Deep dive into Proximal Policy Optimization—the algorithm behind most LLM alignment. Understand trust regions, the clipped objective, GAE, and why PPO's four-model architecture creates problems at scale.
- Reinforcement Learning Foundations for LLM Alignment (opens project or article)
Master the RL fundamentals powering modern LLM training: from MDPs and policy gradients through value functions and actor-critic methods. The mathematical foundations you need before diving into PPO, GRPO, and beyond.
2025
- Deploying Contextual Bandits: Production Guide and Offline Evaluation (opens project or article)
Systems design, offline evaluation, and monitoring strategies for running contextual bandits safely in production.
- Neural Contextual Bandits for High-Dimensional Data (opens project or article)
When linear models fail, neural networks step in. Learn when to use neural bandits, how to quantify uncertainty with bootstrap ensembles, and handle high-dimensional action spaces with embeddings and two-stage selection.
- Implementing Contextual Bandits: Complete Algorithm Guide (opens project or article)
Complete Python implementations of ε-greedy, UCB, LinUCB, and Thompson Sampling. Learn which algorithm to use for your problem with default hyperparameters and practical tuning guidance.
- Contextual Bandit Theory: Regret Bounds and Exploration (opens project or article)
Understand the theory behind contextual bandits: regret bounds, the exploration-exploitation tradeoff, reward models, and why certain algorithms work. Math that directly informs practice.
- When to Use Contextual Bandits: The Decision Framework (opens project or article)
Stop running month-long A/B tests that leave value on the table. Learn when contextual bandits are the right choice for adaptive, personalized optimization—and when to stick with simpler alternatives.
- Beyond the Vibe Check: A Systematic Approach to LLM Evaluation (opens project or article)
Stop relying on gut feelings to evaluate LLM outputs. Learn systematic approaches to build trustworthy evaluation pipelines with measurable metrics, proven methods, and production-ready practices. A practical guide covering faithfulness vs helpfulness, LLM-as-judge techniques, bias mitigation, and continuous monitoring.
2024
-
Became a father
Our first child, Benjamim, arrived on September 12, 2024 — and everything changed in the best possible way. I hit pause on the deep-learning roadmap to start collecting training data from the tiniest (and most fascinating) human dataset I’ll ever work with. These days, midnight diaper shifts feel a lot like reinforcement-learning loops — except the reward signal is a sleepy giggle that makes every iteration worth it. I still catch myself jotting notes in our “family lab notebook,” half scientist, half dad, completely in awe.
2023
- Large Language Models with MLX (opens project or article)
I explored chat tooling on Apple Silicon using MLX to understand the runtime and packaging story.
- LoRA and DoRA Implementation (opens project or article)
I implemented LoRA and DoRA from scratch in PyTorch to understand the methods end to end.
- OpenELM Notes (opens project or article)
I wrote about OpenELM and how Apple approaches efficient language models.
- RAG System with LlamaIndex, Elasticsearch & Llama3 (opens project or article)
A deep dive into building a local-first retrieval-augmented generation system for document Q&A.