Join us

ContentUpdates and recent posts about AIStor..
Link
@faun shared a link, 7 months ago
FAUN.dev()

Jupyter Agents: training LLMs to reason with notebooks

Hugging Face dropped an open pipeline and dataset for training small models—think **Qwen3-4B**—into sharp **Jupyter-native data science agents**. They pulled curated Kaggle notebooks, whipped up synthetic QA pairs, added lightweight **scaffolding**, and went full fine-tune. Net result? A **36% jump .. read more  

Jupyter Agents: training LLMs to reason with notebooks
Link
@faun shared a link, 7 months ago
FAUN.dev()

Building a Natural Language Interface for Apache Pinot with LLM Agents

MiQ plugged **Google’s Agent Development Kit** into their stack to spin up **LLM agents** that turn plain English into clean, validated SQL. These agents speak directly to **Apache Pinot**, firing off real-time queries without the usual parsing pain. Behind the scenes, it’s a slick handoff: NL2SQL .. read more  

Building a Natural Language Interface for Apache Pinot with LLM Agents
Link
@faun shared a link, 7 months ago
FAUN.dev()

Implementing Vector Search from Scratch: A Step-by-Step Tutorial

Search is a fundamental problem in computing, and vector search aims to match meanings rather than exact words. By converting queries and documents into numerical vectors and calculating similarity, vector search retrieves contextually relevant results. In this tutorial, a vector search system is bu.. read more  

Link
@faun shared a link, 7 months ago
FAUN.dev()

5 Free AI Courses from Hugging Face

Hugging Face just rolled out a sharp set of free AI courses. Real topics, real tools—think **AI agents, LLMs, diffusion models, deep RL**, and more. It’s hands-on from the jump, packed with frameworks like LangGraph, Diffusers, and Stable Baselines3. You don’t just read about models—you build ‘em i.. read more  

Link
@faun shared a link, 7 months ago
FAUN.dev()

Inside NVIDIA GPUs: Anatomy of high performance matmul kernels

NVIDIA Hopper packs serious architectural tricks. At the core: **Tensor Memory Accelerator (TMA)**, **tensor cores**, and **swizzling**—the trio behind async, cache-friendly matmul kernels that flirt with peak throughput. But folks aren't stopping at cuBLAS. They're stacking new tactics: **warp-gro.. read more  

Inside NVIDIA GPUs: Anatomy of high performance matmul kernels
Link
@faun shared a link, 7 months ago
FAUN.dev()

Becoming a Research Engineer at a Big LLM Lab - 18 Months of Strategic Career Development

To land a big career role like Mistral, mix efficient **tactical** moves (like LeetCode practice) with **strategic** ups, like building a powerful portfolio and a solid network. Balance is key; aim to impress and prepare well without overlooking the power of strategy in shaping a successful career... read more  

Link
@faun shared a link, 7 months ago
FAUN.dev()

Shai-Hulud npm Supply Chain Attack

Malicious npm packages just leveled up: this one dropped a self-spreading worm that hijacks repos and leaks secrets the moment it lands. It abuses `postinstall` scripts to run TruffleHog and swipe tokens straight from your codebase. Then it uses GitHub Actions to exfiltrate the loot and auto-publis.. read more  

Shai-Hulud npm Supply Chain Attack
Link
@faun shared a link, 7 months ago
FAUN.dev()

How FinOps Drives Value for Every Engineering Dollar

Duolingo’s FinOps crew didn’t just track cloud costs—they wired up sharp, automated observability across 100+ microservices. Real-time alerts now catch AI and infra spend spikes before they torch the budget. They sliced TTS costs by 40% with in-memory caching. Dumped pricey CloudWatch metrics for P.. read more  

How FinOps Drives Value for Every Engineering Dollar
Link
@faun shared a link, 7 months ago
FAUN.dev()

Demystifying Log Retention in Azure

Azure logs come in three flavors: **Activity Logs**, **Diagnostic Logs**, and **Log Analytics**. Each with its own rules for retention and billing. The catch? Those differences aren’t quirks—they’re baked in... read more  

Link
@faun shared a link, 7 months ago
FAUN.dev()

Observability for the Invisible: Tracing Message Drops in Kafka Pipelines

When an event drops silently in a distributed system, it is not a bug, it is an architectural blind spot. Detect, debug, and prevent message loss in Kafka-based streaming pipelines using tools like OpenTelemetry, Fluent Bit, Jaeger, and dead-letter queues. Make sure observability gaps in event strea.. read more  

AIStor is an enterprise-grade, high-performance object storage platform built for modern data workloads such as AI, machine learning, analytics, and large-scale data lakes. It is designed to handle massive datasets with predictable performance, operational simplicity, and hyperscale efficiency, while remaining fully compatible with the Amazon S3 API. AIStor is offered under a commercial license as a subscription-based product.

At its core, AIStor is a software-defined, distributed object store that runs on commodity hardware or in containerized environments like Kubernetes. Rather than being limited to traditional file or block interfaces, it exposes object storage semantics that scale from petabytes to exabytes within a single namespace, enabling consistent, flat addressing of vast datasets. It is engineered to sustain very high throughput and concurrency, with examples of multi-TiB/s read performance on optimized clusters.

AIStor is optimized specifically for AI and data-intensive workloads, where throughput, low latency, and horizontal scalability are critical. It integrates broadly with modern AI and analytics tools, including frameworks such as TensorFlow, PyTorch, Spark, and Iceberg-style table engines, making it suitable as the foundational storage layer for pipelines that demand both performance and consistency.

Security and enterprise readiness are central to AIStor’s design. It includes capabilities like encryption, replication, erasure coding, identity and access controls, immutability, lifecycle management, and operational observability, which are important for mission-critical deployments that must meet compliance and data protection requirements.

AIStor is positioned as a platform that unifies diverse data workloads — from unstructured storage for application data to structured table storage for analytics, as well as AI training and inference datasets — within a consistent object-native architecture. It supports multi-tenant environments and can be deployed across on-premises, cloud, and hybrid infrastructure.