19 stories
On the 17th of August 2026, GitHub had a major global outage. This is a summary of it's incident report. Incident report: https://www.githubstatus.com/incidents/zkxwbgr0cnmx AI Engineering Website: https://aiengg.dev #LLMs #AI #Github
In this live class, we will be going through the "Why Do Multi-Agent LLM Systems Fail?" research paper by Mert Cemri, Melissa Z. Pan and others, with Matei Zaharia and Ion Stoica. The paper studies 1,642 traces from seven multi-agent frameworks and names 14 ways these systems fail.
Let's see where this goes. Incident Report: https://alignment.openai.com/misalignment-reports/an-agent-used-dns-to-reach-an-external-chatbot/ AI Engineering Cohort: https://aiengg.dev #AI #OpenAI #Guardrails
OpenAI breached Hugging Face's production servers. This video summarizes its incident report. Incident: https://openai.com/index/hugging-face-incident-and-the-road-ahead/ AI System Design Cohort: https://aiengg.dev
The AI System Design Cohort starts on 7th November. If you are a software engineer looking to build scalable AI applications, this cohort will be useful. 1. Intro to Agentic System Design 2. Search & Retrieval Patterns 3. Evals & Agentic AI Security 4. Agent Deployment & CI/CD 5.
Most modern Agent Harnesses use Tool Search. This let's them benefit from prompt caching. The LLMs can cache a prefix string in KV cache. Tool search let's Agents reuse a large part of the prefix string.
If our AI chatbot is taking too long to respond, what can we do to reduce latency? This is a real interview question from a big tech company, asked to a candidate in their technical interview round. The video explains the answer in ~6 minutes. To crack more questions, subscribe to the channel!
Jev uses an encoder only architecture, unlike GPT that uses a decoder only architecture. This is a breakdown of how it works internally. References: https://typesafe.ai/blog/introducing-system-one-models-and-jev https://www.langchain.com/blog/building-prod-with-jev-and-langgraph…
If your RAG pipeline is inaccurate, what should you do? This is a real interview question from a big tech company, asked to a candidate in their technical interview round. The video explains the answer in roughly 7 minutes. To crack more questions, subscribe to the channel!
How Software Engineers Stay Relevant in the Age of AI. Moving toward a Forward Deployed Engineer or Solutions Architect role? In this conversation, Arvind and I talk about the fundamentals every engineer should learn, why taste and evaluation matter more than code speed, and how to build…
If you have a 1M Context Window, do you really need anything else? This is a real interview question from a big tech company, asked to a candidate in their technical interview round. The video explains the answer in roughly 7 minutes. To crack more questions, subscribe to the channel!
Join us to know about the AI Engineering Roadmap of 2026. The AI engineering cohort starts on 12th September 2026. A focused, intense 8-week program for software engineers who want real, production-grade AI skills. Website: https://aiengg.dev/ Cheers!
This is called Speculative Decoding. It speeds up LLMs by ~100%. References: OpenAI: https://openai.com/index/gpt-5-6-frontier-intelligence-efficiency/ Deepseek: https://arxiv.org/pdf/2607.05147 Gemini: https://research.google/blog/looking-back-at-speculative-decoding/ AI Engineering Website:…
How do we choose between RAG pipelines and agentic pipelines? This is a real interview question from a big tech company, asked to a candidate in their technical interview round. The video explains the answer in roughly 11 minutes. To crack more questions, subscribe to the channel!
Paper: https://arxiv.org/abs/2309.15217 The AI engineering cohort starts on 12th September 2026. A focused, intense 8-week program for software engineers who want real, production-grade AI skills. 00:00 Paper Reading 26:26 AI Cohort Description Website: https://aiengg.dev/ Cheers!
How do we evaluate RAG pipelines with LLMs? This is a real interview question from a big tech company, asked to a candidate in their technical interview round. The video explains the answer in roughly 11 minutes. To crack more questions, subscribe to the channel!
Over 2000 Indians die every year due to so called "pothole" related accidents. Here is how I used Codex to report them to BBMP. #ChatGPTWork #Sol #Codex
Showing 18 of 19 stories
Load more storiesGet the top AI news and tools straight to your inbox every week.
Weekly · AI news
Join 15,000+ AI enthusiasts
