Tag: Machine Learning
Posts with this tag
- Towards Robust Reinforcement Learning for Small-Scale Language Model Agents: Architecture and Practical Insights
An in-depth analysis of three common PPO failure modes (gradient freezing, numerical overflow, policy collapse) for 70M-500M SLMs, and the Capacity-Headroom Hypothesis (PPL < 20).
- Ornith 1.0 and Self-Scaffolding: Training, Evaluation, and Trust Boundaries for Agentic Coding
A structured look at Ornith-1.0's Self-Scaffolding, anti-reward-hacking controls, and Pipeline-RL design, plus the evaluation and trust boundaries needed for agentic coding systems.
- 24% Conversion Boost! DoorDash Unveils the Underlying Architecture of its Ask Assistant Smart Shopping Agent
An in-depth look at how delivery giant DoorDash combined LLMs, domain-specific AI Agents, Model Context Protocol (MCP), and a three-tier memory system to build an enterprise-grade AI shopping assistant capable of running 2,000 automated evaluations daily.
- What Is Grok 4.5? Capabilities, Benchmarks, and Adoption Decisions
A source-backed review of Grok 4.5's API specifications, coding benchmarks, availability, and the limitations teams should validate before adoption.
- GPT-5.6 Sol Is Generally Available: Routing, Pricing, and Benchmark Caveats
An updated guide to Sol, Terra, and Luna after GPT-5.6 moved from limited preview to general availability, covering API specifications, pricing, benchmark limits, and adoption decisions.
- Ultimate Cloud Computing Liberation: Google Colab CLI Officially Launched, The Strongest Assistant for AI Agents and Developers
Google announces the release of the brand-new Colab CLI tool! Breaking the barrier between local and cloud GPUs, you can instantly invoke powerful computing power through simple terminal commands. It is also the perfect tool for the automated execution of next-generation AI Agents.
- Kaggle Titanic: From 0.74 to 0.816, Feature Engineering Outperforms Parameter Tuning
A complete practical record of the Titanic survival prediction competition: progressive feature engineering, CatBoost and RF ensembling, decoupling CV from Public LB, strict notebook porting, and knowing when to stop. Final Public LB 0.81578.
For speaking invitations, internal engineering sessions, or architecture exchange, see the topics and public work I can bring into the conversation.
Speaking & contact