Digital Factory AI Lead

Reshaping industry workflows by developing AI solutions and effective agentic systems.

Lead AI Engineer

moatless

Helping industry experts and researchers to navigate the sheer endless sea of Large Language Model (LLM) research & innovation because everybody deserves to find the best information.

Founder

Stealth

I was part of a 6 month government-funded Entrepreneurship program that enabled me to work on my own projects and ideas within the LLM space.

During this time, I gained experience in building LLM-powered backends with popular in-demand tools. The prototype I ended up building included the following:

- Fetching data from an external API & processing the response (Python)
- NoSQL database (MongoDB, Postgres)
- PDF Processing (PyPDF2, PyMuPDF)
- Document Extraction (Textract/Document Transformer)
- LLM Chain with multiple steps and prompt templates (Python, Guidance/LangChain)
- API endpoint for the processed data & the LLM outputs (FastAPI)
- Cloud Infrastruture (AWS)

Since I operated on a minimal budget, efficiency and cost reduction were a priority for me. Specifically, I focused on optimizing tasks like PDF processing, document understanding, information extraction and summarization.

Senior Machine Learning Engineer

A1 Telekom Austria AG

During my time as a Senior at A1, I was the DS lead for the company's most advanced NLP projects.

My work included

- Reporting results to several levels of management 
- Representing cutting edge Deep Learning & NLP at the company
- Pitching these topics to the C-Level and other influential exexutives
- Giving expert talks & presentations
- Participating in workshops
- Taking part in the recruiting process for full-time Data Science internships
- Guiding selected interns and assigning them roles in my projects

For a technical overview of my work, see below.

Senior Data Scientist

I was in charge of a Social Media Listening project for Facebook, Twitter and Online Newspapers which included the following

- SQL & MongoDB databases
- Web scraper for the newspaper(s)
- Processing Facebook & Twitter API data
- Transfer Learning & Language Model Finetuning on in-domain texts
- Multiclass and Multilabel Text Classifiers based on Embeddings and Transformers (the original PyTorch implementation of BERT)
- Tableau Dashboard

On top of that, I was developing intents & entities for customer text channels, including e-mail, web search, specialized chatbots (e.g., billing), the aforementioned Social Media data and several smaller Text & Sequence Classification projects.

I then moved on to explore speech channels and became the person in charge of Speech-to-Text at the company, which included 

- Processing short IVR audio recordings
- Processing long(er) call center recordings 
- Labeling pipeline with custom Prodigy scripts
- Developing a subtitle matching algorithm for aligning audio with text & filtering out errorenous subtitles
- Developing Hybrid HMM-TDNN Speech Recognition models with Kaldi
- Switching to (back then freshly released) bleeding edge libraries & models for End-to-end Deep Learning, such as NVIDIA NeMo & Facebook FAIR
- Improving performance on Austrian dialects & in-domain recordings
- Adapting the in-house Classification models mentioned earlier for the text output of the speech channels

In addition, I was able to work on PoCs in the areas of Document Extraction, Document Classification, Entity Extraction & Question Answering.

Data Scientist

Universität Wien

Research in Cognitive & Economic Psychology:

- Dual process theory (Type I/II thinking)
- Metacognition theory
- Experimental assessment of cognitive biases
- Leveraging aggregated mouse cursor movement data

Junior Researcher

Coursera

Data Science Education

As the founder of LLM Watch, my goal is to curate the most relevant Large Language Model (LLM) research and make it accessible to a broad audience. 

I distill complex technical topics into summaries that respect your time. 

Before embarking on this journey, I led Deep Learning R&D at one of Europe's largest telecom companies. There, I spearheaded numerous AI projects with a focus on Natural Language Processing (NLP) and Automatic Speech Recognition (ASR). 

With more than 6 years of experience in AI, I was there before the hype. 

I try to offer a healthy balance between shiny new things and what really matters.

Pascal Biese

Yann LeCun's first big paper after he left Meta. Here's what they did:

For years, training Joint-Embedding Predictive Architectures (JEPAs) has been more art than science. Stop-gradients, teacher-student networks, carefully tuned EMA schedules – a delicate dance of heuristics to prevent representation collapse. But what if there was a principled way?

Self-supervised learning promised to unlock the power of unlabeled data, yet state-of-the-art methods like DINOv2 and I-JEPA rely on complex recipes with dozens of hyperparameters. Each new dataset or architecture requires extensive tuning. Even worse, the training loss often doesn't correlate with downstream performance, making model selection nearly impossible without expensive labeled evaluation.

The core problem: we've lacked a theoretical understanding of what makes good representations. LeJEPA changes this by proving a fundamental result – isotropic Gaussian embeddings uniquely minimize worst-case risk across downstream tasks. 

Tho achieve this, they introduce SIGReg (Sketched Isotropic Gaussian Regularization), a novel objective that enforces this optimal distribution using random projections and characteristic function matching. The beauty lies in its simplicity: linear complexity, bounded gradients, and provably correct by construction.

LeJEPA eliminates stop-gradients and teacher-student networks entirely while achieving competitive performance – 79% on ImageNet-1k with a frozen ViT-H/14. And here's the best part: it works out-of-the-box across 60+ architectures and diverse domains - in-domain pretraining with LeJEPA on small datasets (even just 1,000 samples) outperforms transfer learning from massive frontier models trained on billions of images.

This could fundamentally shift how we think about foundation models. Instead of relying on ever-larger generic models, practitioners in specialized domains can now train competitive models directly on their data with minimal hyperparameter tuning. 

The fact that LeJEPA's training loss actually correlates 99% with downstream performance also solves the critical model selection problem that has plagued SSL research.

What do you think? Is Yann onto something?
↓ 
𝐖𝐚𝐧𝐭 𝐭𝐨 𝐤𝐞𝐞𝐩 𝐮𝐩? Join my newsletter with 50k+ readers and be the first to learn about the latest AI research: llmwatch.com 💡

AI Lead at PwC <​/​> Daily AI highlights for 80k+ experts 📲🤗

LeJEPA: Provable and Scalable Self-Supervised Learning

Microsoft just announced the "Era of Agentic Organization". Here's what that means:

As LLMs scale their reasoning capabilities, we need new techniques: sequential thinking is too slow, and parallel thinking wastes compute by running independent traces that can't learn from each other mid-process. The industry needs a paradigm that combines the adaptivity of sequential reasoning with the speed of parallelization.

For this reason, Microsoft Research developed AsyncThink. AsyncThink is a new approach where a single LLM learns to play two roles simultaneously: an organizer that dynamically structures the reasoning process through Fork and Join operations, and workers that execute sub-queries concurrently. 

Their key contribution is that the model learns via reinforcement learning to optimize its own thinking structure - deciding when to split problems, which sub-queries to pursue, and when to merge intermediate results.

The results? 28% lower inference latency compared to parallel thinking while maintaining higher accuracy on mathematical reasoning. Models trained only on simple countdown problems successfully generalize their asynchronous thinking capabilities to completely unseen domains like Sudoku and complex math problems.

The more tasks AI is supposed to solve within an organization, the more important it becomes to structure all of this in a scalable and controllable way. AI agents need to learn to organize their reasoning like skilled project managers, adaptively coordinating multiple thinking processes based on the problem at hand. 
↓ 
𝐖𝐚𝐧𝐭 𝐭𝐨 𝐤𝐞𝐞𝐩 𝐮𝐩? Join my newsletter with 50k+ readers and be the first to learn about the latest AI research: llmwatch.com 💡

The Era of Agentic Organization

Can we train billion-parameter models without gradients? Oxford researchers just showed it's possible - at 100x the speed.

Evolution Strategies have always been an elegant alternative to backpropagation. They don't need differentiability, handle noisy objectives gracefully, and parallelize beautifully. But there's been a fundamental problem: they simply don't scale to modern model sizes.

University of Oxford's new EGGROLL algorithm mirrors what made LoRA successful for fine-tuning - instead of perturbing all parameters with full-rank matrices, use low-rank approximations. For each population member, sample matrices A and B to form a rank-r perturbation AB^T rather than a full matrix E. This cuts memory from mn to r(m+n) per layer and reduces forward pass compute proportionally.

What makes this work is subtle but powerful: while each perturbation is low-rank, the final parameter update averages across the entire population, yielding a full-rank update when the population is large enough. The researchers prove this approximation converges to full-rank ES at an O(1/r) rate - faster than typical central limit theorem rates due to the symmetry of the problem.

The results validate the approach across three domains. In pure integer pretraining of recurrent language models, EGGROLL enables stable training with population sizes exceeding 250,000 - orders of magnitude larger than prior work. For LLM reasoning tasks, it matches or exceeds GRPO while maximizing hardware utilization. And critically, it achieves near-inference throughput speeds, making gradient-free optimization genuinely practical at scale.

This opens real possibilities for training systems that can't use backpropagation: end-to-end neurosymbolic architectures, models with discrete components, or deployment scenarios where integer-only computation is essential for efficiency.
↓ 
𝐖𝐚𝐧𝐭 𝐭𝐨 𝐤𝐞𝐞𝐩 𝐮𝐩? Join my newsletter with 50k+ readers and be the first to learn about the latest AI research: llmwatch.com 💡

Evolution Strategies at the Hyperscale

We might be systematically underestimating LLMs.

We've been told that reinforcement learning is essential for frontier reasoning capabilities. Massive compute budgets, curated datasets, reward models - the whole apparatus. 

But new research from Harvard University challenges this expensive paradigm with a surprisingly simple question: can we just sample better from base models?

The paper introduces a training-free algorithm that achieves reasoning performance matching or exceeding GRPO (the standard RL method) across mathematics, coding, and science benchmarks. 

The key insight revolves around the "power distribution" - sampling from p^α instead of p. Unlike low-temperature sampling which greedily averages future token likelihoods, power sampling accounts for entire future completion paths. 

This seemingly subtle difference makes a big difference: it naturally upweights tokens with fewer but higher-likelihood futures, essentially planning ahead rather than getting trapped in dead-ends.

Using Metropolis-Hastings MCMC to approximate this distribution, the researchers show that Qwen2.5-Math-7B can reach 74.8% on MATH500 without any additional training - nearly matching the 78.5% from full RL posttraining. 

On out-of-domain tasks like HumanEval, it outperforms by up to 20 percentage points. Perhaps most remarkably, unlike RL which suffers from diversity collapse, this method maintains strong pass@k performance across multiple samples.

The takeaway is that base models may already contain the reasoning capabilities we're spending enormous resources to extract through RL. This would open up a new axis for inference-time compute scaling that's accessible to anyone with a base model - no training infrastructure, curated datasets, or verifiers required.

We're still learning how to properly query the "intelligence" that already exists in our models.
↓ 
𝐖𝐚𝐧𝐭 𝐭𝐨 𝐤𝐞𝐞𝐩 𝐮𝐩? Join my newsletter with 50k+ readers and be the first to learn about the latest AI research: llmwatch.com 💡

Your Base Model is Smarter Than You Think

IndiaLends

Developed and optimized 13 product funnels in SQL for personal loans, identifying key drop-offs and achieving a 15% increase in offer selection rate.

Worked directly with Product Managers, partnering with developers and tech teams to troubleshoot and resolve API issues, reducing rejections by 17%. Supported backend improvements for CRM automation, boosting accuracy by 20%.

Conducted extensive testing on 30+ APIs with Postman, resolving inconsistencies for 5 major lender integrations and documenting processes to streamline troubleshooting by 40%.

Built 8 Clevertap funnels to analyze offer page interactions, improving offer selection by 8%.

Maintained MIS reports with 99% accuracy, coordinating with Growth and Tech teams to resolve data inconsistencies.

Analyzed 60000+ loan disbursals, identifying top rejection reasons and advising on updates that boosted disbursals by 12%.

Consolidated 15+ lender API data sources, reducing data retrieval time by 30% for efficient reporting.

Product Analyst Intern

Leadle

• Wrote SQL queries to extract data from databases efficiently. Developed, maintained, and optimized SQL Queries and stored procedures.
• Collaborated with RevOps team and maintained live prospects dashboard on Google Looker Studio, connected with Google Sheets
• Prepared and cleaned data for analysis, including data normalization, aggregation, and transformation. (SQL, Excel and Tableau)
• Ensured accurate data by performing thorough validation and quality checks. Worked together with business and data teams to comprehensively grasp the data requirements and objectives.
• Leveraged the power of data visualization using sophisticated tools like Tableau to create captivating and informative reports and dashboards. Presented the findings to stakeholders in a visually engaging and interactive manner, unleashing the full potential of data-driven insights. Translated complex data into clear and understandable visual representations.

Data Analyst

5X your AI & Data Science skills. Get free AI guides and resources ⬇️

I like writing about Data. 

Working as a Product Analyst in a Fintech startup.

I've started writing on LinkedIn from 2023 as part of my job search. 

(Then it became addictive!)

This platform became a powerful aid in my growth journey.

I'm fortunate to meet some amazing people via LinkedIn.

I write about my learnings and opinions on Data career.

Enquiries/Collab: contact.leonjose@gmail.com

Leon Jose

Stanford’s AI Paper Everyone’s Talking About.
You’ll wish you read this sooner.

Have you ever tried to understand how Artificial Intelligence really works?

It sounds exciting, but most explanations are too complex.

Many people know the terms - AI, ML, Deep Learning.

but not what they truly mean..

Stanford recently published a paper that explains it in simple words.

It shows how machines learn, make choices, and improve with data.

No complex math. No jargon. Just clarity.

It’s based on the same Stanford ideas but written for beginners.

Steve makes it easy to see how AI actually works in the real world.

You’ll learn:
➙ What AI and ML really mean
➙ How they work together
➙ Why every company is using them
➙ How you can apply this knowledge in your own work

Read it before everyone else does.

👉 Follow gotspeed to grow your AI skills
♻️ Repost the guide


#ai #stanford

Product @SaveIN | AI | FinTech | Analytics

AI Explained - Stanford Research Paper

Only 1% are building these projects
20 ideas that can change your AI career

If you’re trying to break into the world of AI tools, agents, and automation
you don’t need another course.

That’s what separates people who just consume AI from those who actually create with AI.

But here’s the problem.. most beginners don’t know where to start.

I’ve found an amazing guide by Bhavishya Pandit for Gen AI projects.

It’s a curated list of hands-on project ideas designed to help you:
Learn by doing
Build your first AI-based product
Strengthen your portfolio
And stand out in the AI job market

Each project idea is simple enough to start today.


👉 Follow gotspeed to grow your AI skills
♻️ Repost the guide

#GenAI #ai #interview

GenAI Projects

15 GitHub repos you never saw coming.
Before they disappear into private access.

Building AI agents is exciting.
But it’s also overwhelming.

Where do I start?
How do I find reliable resources?

Luckily, I've found this guide by Om Nalinde with 15 essential GitHub repositories.

Each repository will help you:
➙ Understand how AI agents think and act
➙ Build your first working prototype
➙ Experiment with frameworks like LangChain, AutoGPT, and CrewAI
➙ Learn how tools like memory, planning, and retrieval really work

If you’ve ever wanted to build your own AI agent,
this list will save you weeks of confusion and trial-and-error.

I’ve attached the full guide in this post.

Pick one repository today and start building.

Small consistent progress compounds faster than you think.

👉 Follow gotspeed to grow your AI skills
♻️ Repost the guide


#ai #github #aiagent

Best Github Repos for AI Agents

99% will ignore this AI shift.
But the 1% will build the future.

Google just released a new guide called 𝗔𝗜 𝗔𝗴𝗲𝗻𝘁𝘀.

It’s a practical playbook showing how AI agents are about to reshape how we work.

If you’ve been seeing terms like 𝘈𝘐 𝘈𝘨𝘦𝘯𝘵𝘴, 𝘈𝘶𝘵𝘰𝘯𝘰𝘮𝘰𝘶𝘴 𝘈𝘐, or 𝘞𝘰𝘳𝘬𝘧𝘭𝘰𝘸 𝘈𝘶𝘵𝘰𝘮𝘢𝘵𝘪𝘰𝘯..

this guide is your shortcut to understanding it.

Here’s what you’ll learn inside:
⟶ What AI Agents actually are 
⟶ How they differ from chatbots and copilots
⟶ Real-world examples of how Google is using them
⟶ How developers and product folks can start building them
⟶ The emerging ecosystem of tools and frameworks

👉 Follow gotspeed to grow your AI skills
♻️ Repost the guide


#aiagents #ai

AI Agents Guide for Beginners by Google

They know how to use ChatGPT 
But not how it actually works

Over the past three years, millions started using AI tools daily. 

But only a few understand the foundations that make them possible. 

Knowing this difference is what separates users from builders.

I recently studied a brilliant paper 𝗙𝗼𝘂𝗻𝗱𝗮𝘁𝗶𝗼𝗻𝘀 𝗼𝗳 𝗟𝗮𝗿𝗴𝗲 𝗟𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗠𝗼𝗱𝗲𝗹𝘀 by 𝘛𝘰𝘯𝘺 𝘟𝘪𝘢𝘰 and 𝘑𝘪𝘯𝘨𝘣𝘰 𝘡𝘩𝘶.

Here’s a quick overview before you dive in:
↳ LLMs predict words to learn meaning.
↳ They train on massive text data.
↳ Transformers use attention to find context.
↳ Layers help models reason like humans.
↳ Knowing this helps you build, not just use.

This paper will help you:
✦ Grasp key LLM ideas without jargon
✦ Understand how models like ChatGPT learn
✦ Build a solid base for advanced AI learning

Start learning what most people skip. 

👉 Follow gotspeed to grow your AI skills
♻️ Repost the guide


#ai #llm

Foundations of LLMs

They’re wasting millions on LLMs.
You don’t have to.

Building with LLMs is exciting, until the bill hits.

Most teams start small but end up paying 5x more than they should.

Why? Because they optimize features, not costs.

Bhavishya Pandit created a free PDF guide that breaks down how to cut your model expenses without losing performance.

Here’s what you’ll learn inside:
⟶ How to use caching to save 40% instantly
⟶ Why token limits can silently drain your budget
⟶ The right balance between model precision and cost
⟶ Smart batching strategies for high-volume prompts
⟶ How prompt compression and distillation reduce usage
⟶ And 10+ practical methods real teams use to stay lean

These strategies aren’t theory. they’re used by developers running real AI products.

👉 Follow gotspeed to grow your AI skills
♻️ Repost the guide


#ai #llm

taplio

The Battle

Pascal Biese

Leon Jose

Followers

Posts

Engagements

Likes

Performance ScorePascal Biese's LinkedIn Performance

Performance ScoreLeon Jose's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Pascal Biese

Leon Jose

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);}Pascal Biese's LinkedIn Performance

Performance ScoreLeon Jose's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScorePascal Biese's LinkedIn Performance