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.

AlphaSense

AlphaSense is how the world’s most sophisticated companies remove uncertainty from their decision-making. With market intelligence and search built on proven AI, AlphaSense quickly delivers relevant insights from a trustworthy universe of public and private content—including equity research, company filings, event transcripts, expert calls, news, trade journals, and clients’ own research content.

Regional Product Specialist - France & Benelux

Dataiku

Dataiku is the platform for Everyday AI, systemizing the use of data for exceptional business results. Organizations that use Dataiku elevate their people (whether technical and working in code or on the business side and low- or no-code) to extraordinary, arming them with the ability to make better day-to-day decisions with data.

More than 500 companies worldwide use Dataiku to systemize their use of data, analytics, and AI, driving diverse use cases from fraud detection to customer churn prevention, predictive maintenance to supply chain optimization, and everything in between.

Account Executive

Tessian

Tessian’s mission is to secure the human layer. Using machine learning technology, Tessian automatically stops data breaches and security threats caused by human error - like data exfiltration, accidental data loss, business email compromise and phishing attacks - with minimal disruption to employees' workflow. As a result, employees are empowered to do their best work, without security getting in their way. Founded in 2013, Tessian is backed by renowned investors like Sequoia, Accel and Balderton and has offices in San Francisco and London.

OneTrust

OneTrust is the #1 most widely used privacy, security and trust technology platform used by more than 5,000 companies to comply with the CCPA, GDPR, ISO27001 and hundreds of the world’s privacy and security laws. OneTrust’s primary offerings include OneTrust Privacy Management, OneTrust PreferenceChoice™ consent and preference management, OneTrust Vendorpedia™ third-party risk management, OneTrust GRC integrated risk management and OneTrust Ethics compliance and ethics software.

Account Executive 

Hosco

Hosco was founded in 2011 with a very clear vision: to gather, inspire and offer job opportunities to all professionals in the hospitality industry worldwide. Since then, we have developed a unique concept that connects qualified professionals, employers and hotel schools around the world in a single network.

#EmpoweringHospitality

Sales Executive 

Engineering experiences that amaze. We create technology with a purpose: to make life better for everyone, everywhere.  Keep Reinventing.

Sales Executive

Passionate about tech innovation, I bring 5+ years of experience in data privacy, cybersecurity, and AI/ML platforms.

Transitioning from a dynamic Account Executive role to explore new horizons in Product. 

Excited to blend my diverse experiences and skills into this new journey. With a foundation in strategic sales, I'm ready to leverage my customer-centric approach, critical thinking, and problem-solving abilities for innovative product development. 

Always eager to connect with industry experts and learn from their experiences.  🚀

taplio

The Battle

Andriy Burkov

Fiona C.

Followers

Posts

Engagements

Likes

Performance ScoreAndriy Burkov's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Andriy Burkov

Fiona C.

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