Read my just released 📖 The Hundred-Page Language Models Book: https://thelmbook.com

Read my bestselling 📖 The Hundred-Page Machine Learning Book: https://themlbook.com

Read 📖 Machine Learning Engineering: https://mlebook.com

Subscribe to my weekly ✉️ Artificial Intelligence newsletter: https://www.linkedin.com/newsletters/artificial-intelligence-6598352935271358464/ and to my weekly ✉️ Data Science newsletter: https://www.linkedin.com/newsletters/7102511020270608384/

About me:

Ph.D. in Artificial Intelligence, passionate about data, fluent in English, French, and Russian. Solid scientific programming and team leadership skills, with over 20 years of experience working on various computing projects, including several of my own startups.

More than 15 years of hands-on experience in automated data analysis, machine learning, and natural language processing.

Trained a Transformer from scratch and fine-tuned pretrained transformers for various tasks. Built a robot that crawls the internet, finds websites with business-critical information, and retrieves updated information periodically. Developed an enterprise chatbot that doesn’t hallucinate.

Expert in Python and Java with several years of daily design and development experience in big data contexts.

Specialties: machine learning, natural language processing, conversational interfaces (chatbots), information retrieval.

Andriy Burkov

LLMs process text from left to right — each token can only look back at what came before it, never forward. This means that when you write a long prompt with context at the beginning and a question at the end, the model answers the question having "seen" the context, but the context tokens were generated without any awareness of what question was coming. This asymmetry is a basic structural property of how these models work.

The paper asks what happens if you just send the prompt twice in a row, so that every part of the input gets a second pass where it can attend to every other part. The answer is that accuracy goes up across seven different benchmarks and seven different models (from the Gemini, ChatGPT, Claude, and DeepSeek series of LLMs), with no increase in the length of the model's output and no meaningful increase in response time — because processing the input is done in parallel by the hardware anyway.

There are no new losses to compute, no finetuning, no clever prompt engineering beyond the repetition itself. 

The gap between this technique and doing nothing is sometimes small, sometimes large (one model went from 21% to 97% on a task involving finding a name in a list). If you are thinking about how to get better results from these models without paying for longer outputs or slower responses, that's a fairly concrete and low-effort finding.

Read with AI tutor: https://lnkd.in/ene242cx

Get the PDF: https://lnkd.in/e9tbUTNv

PhD in AI, author of 📖 The Hundred-Page Language Models Book and 📖 The Hundred-Page Machine Learning Book

This paper discusses methods to extend LLMs into interactive agentic systems that incorporate planning, tool use, and search.

The authors from Meta, Amazon, Deepmind, and other top private and public labs consider in-context approaches that adapt during inference without altering weights and post-training methods that refine the model through reinforcement learning.

Read with an AI tutor: https://lnkd.in/efJShRmH

Read alone: https://lnkd.in/ecBVandw

Spent some time with RedisVL recently and it's a surprisingly practical tool. It's a Python library that lets you use Redis as a vector database with such features as semantic search, LLM caching, and chat memory. What I like is that it builds on redis-py, so it's not some new ecosystem to learn. If your team already runs Redis, you can add vector search and retrieval to your AI pipeline without introducing another database. Fewer moving parts is always a win.

Here's the link to the docs: https://fandf.co/4qixIzY

Disclaimer: Redis was last week's sponsor of my weekly AI newsletter.

𝑺𝑽𝑷 - 𝑯𝒆𝒂𝒅 𝒐𝒇 𝑺𝒕𝒓𝒂𝒕𝒆𝒈𝒚 𝑳𝑯 𝑼𝑺𝑨 @ 𝑳𝒂𝒘𝒓𝒆𝒏𝒄𝒆 𝑯𝒂𝒓𝒗𝒆𝒚 | 𝑻𝒆𝒄𝒉𝒏𝒊𝒄𝒂𝒍 𝑹𝒆𝒄𝒓𝒖𝒊𝒕𝒎𝒆𝒏𝒕

Recent Speaking Engagements: 

-	Organization: The AI Summit, 11-Dec-2024 - Topic: How to Recruit & Retain Verified AI Talent. 2:05 PM to 2:15 PM, Next Generation Stage
-	Organization: The AI Summit, 11-Dec-2025 - Topic: Panel: Reskilling for the Economy - From Prompt Engineer to AI Leader. 1:35 PM to 2:15 PM, The AI Playground
-	Organization: The AI Summit, 11-Dec-2025 - Topic: The Next Decade of Scaling AI: Resilient, Responsible, and Transformative. 3:35 PM to 4:15 PM, AI at Scale Stage
-	Organization: The AI Summit, December 2023- Topic: Assessing the Impact of AI on the Future of Work and Jobs. 500+ in attendance at the Javits Center 
-	Organization: The AI Summit, December 2023- Topic: Strategies to Recruit and Retain a Balanced Team of AI Talent. 500+ in attendance at the Javits Center 
-	Organization: DataCamp: Radar Summit, January 2023- Topic: How Data Leaders Can Build an Impactful Data Team in 2023. 9,000 in attendance on a Video Meeting Platform
-	Organization: The AI Summit Summit, December 2022- Topic: How do you Build Your Team Early. 500+ in attendance at the Javits Center 
-	Organization: DataCamp: Radar Summit, July 2022- Topic: Building and Scaling Remote Data Teams. 18,000 in attendance on Zoom
-	Organization: LHi: Leadership Summit, April 2022- Topic: Future You. 20 in attendance, in person 
-	Organization: DataCamp: Data Executive Meetup, December 2021- Topic: Building High Impact Data Teams- 70 in attendance on the USS Intrepid
-	Organization: LHi: Leadership Summit,  February 2020- Topic: The Journey. 20 in attendance, in person

I am passionate about helping businesses find and retain the best talent in the market. As the SVP - Head of Strategy LH USA, I oversee the strategic direction and growth of Lawrence Harvey's US operations, focusing on expanding our client base, diversifying our service offerings, and enhancing our brand reputation.

I joined Lawrence Harvey in 2015 as an associate recruitment consultant and quickly rose through the ranks to become the head of the data and AI practice, working with blue-chip organizations across various sectors. I have built and led a high-performing team of recruiters and managers, delivering exceptional results and customer satisfaction. I have also contributed to the development and implementation of innovative solutions and best practices, leveraging my expertise in analytics and data-driven decision making. I am always looking for new opportunities and challenges to grow personally and professionally, and to make a positive impact in the tech recruitment industry.

taplio

The Battle

Andriy Burkov

Mariano Allegra

Followers

Posts

Engagements

Likes

Performance ScoreAndriy Burkov's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Andriy Burkov

Mariano Allegra

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Andriy Burkov's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreAndriy Burkov's LinkedIn Performance