SCC Comets

$Domain-Specialized Large Language Models: Code, Math, and Medicine$

Tess Rempel Mar, 15 2026 Artificial Intelligence

Domain-Specialized Large Language Models: Code, Math, and Medicine

Domain-specialized LLMs like CodeLlama, Med-PaLM 2, and MathGLM are outperforming general AI models in code, math, and medicine - with higher accuracy, lower costs, and real-world impact.

Tess Rempel Mar, 13 2026 Artificial Intelligence

Autoscaling Large Language Model Services: Policies, Signals, and Costs

Autoscaling LLM services requires specialized metrics like prefill queue size and slots_used-not CPU or GPU usage. Learn how to balance cost and latency with real-world policies for chatbots, batch jobs, and real-time apps.

Tess Rempel Mar, 11 2026 Artificial Intelligence

Multilingual RAG for Large Language Models: Overcoming Cross-Language Retrieval Challenges

Multilingual RAG lets LLMs answer questions across languages using external data-but hidden biases favor English. Learn how cross-language retrieval fails, why translation alone isn’t enough, and what new frameworks like DKM-RAG are doing to fix it.

Tess Rempel Mar, 7 2026 Artificial Intelligence

Vision-Language Applications with Multimodal Large Language Models

Vision-language models now combine image and text understanding in real-time, transforming document processing, healthcare, and robotics. Open-source models like GLM-4.6V are outperforming proprietary systems in key areas, despite high compute costs and hallucination risks.

Tess Rempel Mar, 6 2026 Artificial Intelligence

Token Budgets and Quotas: How to Stop LLM Cost Overruns Before They Happen

Token budgets and quotas are the only way to prevent runaway LLM costs. Learn how to set limits on token usage, apply graduated throttling, and avoid expensive mistakes before they happen.

Tess Rempel Mar, 5 2026 Artificial Intelligence

How Large Language Models Use Probabilities to Choose Words and Phrases

Large language models generate text by calculating the probability of each next word based on patterns learned from massive datasets. This article explains how they use probabilities, decoding strategies like top-p and temperature, and why they sometimes make mistakes - even when they sound convincing.

Tess Rempel Mar, 4 2026 Artificial Intelligence

How to Set Realistic Expectations for Vibe Coding on Enterprise Projects

Vibe coding isn't magic - it's a new way to build software using AI as a proactive team member. Learn how to set realistic expectations, write clear specs, and avoid common enterprise pitfalls that derail adoption.

Tess Rempel Mar, 3 2026 Artificial Intelligence

Why Tokenization Still Matters in the Age of Large Language Models

Tokenization remains critical in the age of large language models, impacting cost, accuracy, and efficiency. Learn why subword tokenization, vocabulary size, and domain-specific tuning still make or break LLM performance.

Tess Rempel Mar, 2 2026 Artificial Intelligence

Continuous Improvement Loops: How Feedback, Retraining, and Prompt Updates Keep AI Models Accurate

Continuous improvement loops keep AI models accurate by using real-world feedback, automated retraining, and prompt updates. Without them, models degrade quickly. Here’s how to build one that actually works.

Tess Rempel Mar, 1 2026 Artificial Intelligence

Abstention Policies for Generative AI: When the Model Should Say It Does Not Know

Generative AI often invents answers instead of admitting ignorance. Learn how abstention policies teach models to say 'I don't know'-and why that’s the key to trustworthy AI.

Tess Rempel Feb, 28 2026 Artificial Intelligence

Teaching with Vibe Coding: Learn Software Architecture by Inspecting AI-Generated Code

Learn software architecture by inspecting AI-generated code instead of writing it first. Vibe coding flips traditional programming education, helping students understand system design before syntax. Used by Google, Amazon, and top bootcamps.

Tess Rempel Feb, 27 2026 Artificial Intelligence

Data Collection and Cleaning for Large Language Model Pretraining at Web Scale

Training large language models requires cleaning hundreds of terabytes of web data. Discover how top teams filter, deduplicate, and curate data to boost performance - and why quality matters more than quantity.