Blog
Latest articles
7 min readStableLM 2: Stability AI's Compact 1.6B Model for Edge Inference
StableLM 2 1.6B outperforms Phi-1.5 and TinyLlama at its size class and is small enough to run on a Raspberry Pi, in a browser via WebLLM, or on old consumer hardware.
9 min readSemantic Search and Entities: Designing Websites for Google Knowledge Graph
Modern search engines index concepts and connections, not just keywords. Learn how to optimize your markup for entity relationship crawls.
7 min readAmazon Nova: AWS's Native Foundation Models on Bedrock
Amazon Nova offers three tiers - Micro, Lite, and Pro - with up to 300k context on Nova Pro, multimodal input, and deep AWS ecosystem integration via Bedrock.
7 min readMLflow 2.x for LLMs: Track Prompts, Responses, and Fine-Tune Runs
MLflow 2.x adds native LLM tracing, prompt versioning, and model registry support - bringing the same experiment discipline from ML training to LLM application development.
9 min readThe Needle-in-a-Haystack Test: Benchmarking LLM Long-Context Recall
The needle-in-a-haystack test measures whether an LLM can recall a single specific fact embedded at varying depths in a long document, revealing which models have uniform long-context recall and which have blind spots.
7 min readEvidently AI: Monitor ML Models for Data Drift and Performance Degradation
Evidently AI generates data drift reports, quality checks, and model performance dashboards for production ML - catching distribution shifts before they silently corrupt your predictions.
8 min readText Generation Inference (TGI): HuggingFace's Production LLM Server
TGI is HuggingFace's open-source LLM serving engine with continuous batching, tensor parallelism, and an OpenAI-compatible API - deployable in one Docker command.
4 min readZustand 5: The Minimal React State Manager That Scales
Zustand 5 tightens TypeScript types, removes the act() auto-wrap, and adds stricter equality — here is how to use it effectively from simple stores to large sliced architectures.
8 min readRanking in Perplexity AI Search: The New SEO Frontier in 2026
Perplexity is processing over 1.5 million queries daily and growing fast. Getting cited there requires a different optimization strategy than Google SEO.
7 min readGroq LPU: How to Get 800+ Tokens/sec LLM Inference
Groq's Language Processing Unit achieves 800+ tokens/sec on Llama 3.1 70B - 10-20x faster than GPU inference. Here's how to use GroqCloud and integrate it into existing OpenAI pipelines.
7 min readBLIP-2: Bootstrap Vision-Language Models With Frozen Image Encoders
BLIP-2 bridges a frozen CLIP image encoder and a frozen LLM through a lightweight Q-Former, achieving strong VQA and captioning performance without updating the large pretrained components.
7 min readGemma 3 27B: Google's Multimodal Open Model That Surpasses Llama 3.3
Gemma 3 27B scores 67.5% on MMLU while adding native image understanding and 128k context - outperforming Llama 3.3 70B on MMLU at less than half the parameter count.