I share my learning journey into Data Science with my amazing LinkedIn friends, follow me and let's grow together!

For downloadable ML/ AI/ Data Science learning materials, please check my previous posts.

** Opinions are my own and not the views of my employer **

Alex Wang

Data Scientist

GenAI.Works

Driving the strategic vision for education at GenAI.works, working with teams to develop innovative AI-powered learning initiatives to empower professionals and align with the evolving needs of the workforce.

Head of Education Strategy

AI4Diversity

AI 4 Diversity (AI4D) is a nonprofit global initiative that engages and educates diverse communities about AI to benefit global society.

Member of Operation Team

Quantium

Global Market - ML Algorithms

Deloitte Australia

Business Algorithms

University of Technology Sydney

Courses: Advanced Machine Learning, ML Fundamentals

Casual Academic

Hackmakers

Pactera

Data Analyst

I share my learning journey into AI & Data Science here, click 'follow' and let's grow together!

For open access resources, productive tools, and learning materials in AI/ ML/ Data Science, please check my previous posts.

** Opinions are my own and not the views of my employer **

Member of Global Committee

It is voluntary role helping the global management team. 
AI 4 Diversity (AI4D) is a nonprofit global initiative that engages and educates diverse communities about AI to benefit global society.

AI and information-processing technologies are expected to transform 86% of businesses - more than any other category.

The World Economic Forum’s recent report shows how quickly things are shifting:
▪️By 2030, 39% of today’s core skills will be outdated, and 59% of workers will need reskilling.

▪️And alongside that, roughly 170 million jobs will be created, while 92 million will disappear.

It’s a lot of change happening at once - and AI sits right in the middle of it.

If you’re thinking about upskilling, Coursera is offering 40% off Coursera Plus Annual until 12/1 🔥

Check it out before ends
https://lnkd.in/gGuZTv7g

Here are some practical AI courses for professionals (all available on Coursera):

1----- AI for everyone from DeepLearning AI
 imp.i384100.net/Wy1zaO

2----- AI Essentials Specialization from Google
 https://lnkd.in/gGDFt3Uw

3----- Prompting Essentials Specialization from Google
 https://lnkd.in/gEWQeYSv

4----- Building with LLMs from DeepLearning AI
 https://lnkd.in/gh3qcVGY

5----- IBM AI Developer
 https://lnkd.in/g_bMxGVY

6----- GenAI Engineering from IBM
https://lnkd.in/g5xCkqj9

7----- Prompt Engineering Basics from IBM
https://lnkd.in/g-TxdUvV

Happy learning!
__________

For more on AI and learning materials, plz check my previous posts. 

I share my journey here. Join me and let's grow together. Alex Wang

#artificialintelligence #technology #machinelearning #datascience #business

Learn AI Together - I share my learning journey into AI & Data Science here, 90% buzzword-free. Follow me and let's grow together!

McKinsey's latest AI report 2025 - 
the gap between “using AI” and “getting value from AI” is bigger than most think.

A few highlights + some thoughts:

𝟭) 𝗔𝗜 𝗶𝘀 𝗲𝘃𝗲𝗿𝘆𝘄𝗵𝗲𝗿𝗲, 𝗯𝘂𝘁 𝘁𝗿𝗮𝗻𝘀𝗳𝗼𝗿𝗺𝗮𝘁𝗶𝗼𝗻 𝗶𝘀𝗻’𝘁

88% of organisations say they use AI somewhere…
…but nearly two-thirds are still stuck in pilots.
Only one-third have scaled AI across the business.

This matches what I see with teams: the tech is ready, but workflows, processes, and culture usually aren’t.

𝟮) 𝗔𝗜 𝗮𝗴𝗲𝗻𝘁𝘀 𝗮𝗿𝗲 𝗿𝗶𝘀𝗶𝗻𝗴, 𝗯𝘂𝘁 𝘀𝘁𝗶𝗹𝗹 𝗲𝗮𝗿𝗹𝘆

62% are experimenting with agents.
Only 23% have reached any level of scaling.
And usually in just one or two functions.

Lots of buzz, but the reality on the ground is: most orgs are still figuring out what “agentic AI” actually means in practice.

𝟯) 𝗧𝗵𝗲 𝗿𝗲𝗮𝗹 𝘃𝗮𝗹𝘂𝗲 𝗶𝘀 𝘀𝗵𝗼𝘄𝗶𝗻𝗴 𝘂𝗽 𝗶𝗻 𝗽𝗼𝗰𝗸𝗲𝘁𝘀

Innovation ↑
Customer satisfaction ↑
Engineering & ops costs ↓

But only 39% report any EBIT impact, and most say it’s <5%.

𝟰) 𝗧𝗵𝗲 𝘁𝗼𝗽 𝟲% (𝗠𝗰𝗞𝗶𝗻𝘀𝗲𝘆 𝗰𝗮𝗹𝗹𝘀 𝘁𝗵𝗲𝗺 “𝗔𝗜 𝗵𝗶𝗴𝗵 𝗽𝗲𝗿𝗳𝗼𝗿𝗺𝗲𝗿𝘀”) 𝗹𝗼𝗼𝗸 𝗰𝗼𝗺𝗽𝗹𝗲𝘁𝗲𝗹𝘆 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁

What separates them?
Not models. Not tools. Not vendors.

They:

- Set bold ambitions (not just efficiency, but growth + new business models)
- Redesign workflows end-to-end
- Have leaders deeply involved, not just sponsoring from a distance
- Invest meaningfully (often >20% of digital budget goes into AI)
- Scale agents 3–5× faster across functions

I’ve seen this too: tech matters, but ambition, workflow, and ownership matter more.

𝟱) 𝗧𝗵𝗲 𝗻𝗲𝘅𝘁 𝟭𝟮 𝗺𝗼𝗻𝘁𝗵𝘀 𝘄𝗶𝗹𝗹 𝗯𝗲 𝗺𝗲𝘀𝘀𝘆

More orgs expect workforce shifts, both reductions and new hiring, especially for AI-adjacent skills like data engineering, MLOps, and product.

Generative AI adoption is now mainstream.
Agentic AI is on the rise.
But true enterprise transformation has barely started.

2023–24 was about experimenting.
2025–26 will be about rewiring.

Drop link in the post. 
__________

For more on AI and learning materials, plz check my previous posts.

I share my journey here. Join me and let's grow together. Alex Wang

#bussiness #technology #generativeai #aiagents

The state of AI in 2025

Google is on a roll. After Gemini 3, Google just dropped an AI idea that challenges scaling itself.

💡Overestimated or the real deal?

Let’s start with what it’s about (and why it might be one of Google’s most interesting theoretical advances in years):

▪️This paper challenges a core assumption in deep learning:
that stacking more layers and scaling models is the main path to intelligence.

▪️Instead, Google researchers propose Nested Learning - a paradigm where neural networks are treated as systems of nested optimization problems, each with its own memory, update frequency, and context flow.

They also introduce a prototype architecture called HOPE, built around these ideas to improve memory, continual learning, and adaptability.

It’s a rare paper that asks:
“How should models learn, not just how can they get bigger?”

🔗Btw, an awesome live workshop on Accelerating JSON Array Queries is coming soon; hands-on, free to join

Details here: https://lnkd.in/eQpFiD6E

*Hosted by SingleStore, could be helpful useful if you work with structured data. 
----------
✔️My take:

1) --- An exciting theoretical direction; super valuable for research, but not production-ready yet.

Nested Learning offers a clearer way to understand in-context learning and model adaptation - closer to how humans process and store knowledge.

But current results are on small-scale benchmarks.
We don’t yet know how it performs at real-world scale or under production demands.

2) --- A glimpse into the “post-Transformer” era

Multi-timescale memory, self-modifying components, optimizers as slow knowledge, these ideas may influence future architectures.

3) --- Is it worth watching? 𝗗𝗲𝗳𝗶𝗻𝗶𝘁𝗲𝗹𝘆

It’s not a replacement for the LLMs we use today,
but it’s an early hint at what might come after scaling.

Image: Google page, Pete Reese
__________

For more on AI and learning materials, plz check my previous posts. 

I share my journey here. Join me and let's grow together. Alex Wang

#Technology #AI #DeepLearning #LLMs #Transformers

Building AI agents = 5% AI + 95% software engineering
Not exactly accurate, but you get the point.

We don’t really build AI agents - we architect systems, and let the AI sit inside them.

This diagram from Microsoft Azure shows what real, enterprise-ready AI infrastructure looks like.

It’s software architecture, system design, and reliability engineering:

✅ Identity management
✅ Document filtering (ACLs, redaction, PII masking)
✅ Schema mapping
✅ Human-in-the-loop escalation
✅ Infra that scales across both vector + SQL

Think of AI agents like APIs that reason, not magic.
They still need:

- Fine-grained access control
- Separation of structured vs unstructured knowledge
- Observability, fallback routing, and data traceability
- Flows connecting document pipelines with model orchestration

That’s why I find tools like Zapier interesting right now -
They’re quietly turning automation flows into orchestration layers for agents.
Connecting APIs, reasoning steps, and human review - all without code.

If you want to experiment with it yourself, try it here:
 🔗 https://lnkd.in/gKpEGnNG 

Before you tune the prompts, build the foundations.
__________

For more on AI and learning materials, plz check my previous posts.

I share my journey here. Join me and let's grow together. Alex Wang

#aiagents #agetnicai #llms #artificialintelligence

"Which AI role suits you?"

An interesting matrix, and I like the idea.

It might simplify reality a bit, but it’s directionally accurate and still useful as a reference to plan your learning path and get a quick sense of:
- What AI roles exist
- What core skills each role requires

If you're aiming for a specific role, this helps you understand what’s critical vs “nice to have.”

Here are a few thoughts I’d add, especially from a GenAI lens (understandably hard to cover all in one graph):

▪️In the matrix, LLMs, RAG, and Agents are grouped as one row -
but in practice, these are completely different skill sets (imo):

→ Prompting ≠ Retrieval-Augmented Generation ≠ Multi-Agent Orchestration
(And most teams today need all three)

▪️Worth noting a few other key skills that often come into play in production GenAI systems:
→ Evaluation and observability
→ Guardrails, tool-calling, tracing
→ Feedback loops, context management, orchestration logic

▪️Business roles are much more dimensional in practice -
For example, AI PMs and Ethics leads often need deep understanding of compliance, risk, model lifecycle… not just “soft skills.”

📍Also, there’s an awesome 2‑day AI camp happening this weekend, hosted by Outskill.
Upskilling‑oriented | 16 hours | free to join:
🔗 https://lnkd.in/gBuXuqJ7

Curious:
Does this reflect your current AI role?
Any skills you’d add or move?

Image: SuperDataScience
__________

For more on AI and learning materials, plz check my previous posts. 

I share my journey here. Join me and let's grow together. Alex Wang

#artificialintelligence #machinelearning #aiagents #generativeai

One of the clearest explanations I’ve seen on how transformers actually work.

LLMs like ChatGPT are built on this breakthrough architecture - the idea that unlocked modern AI.

This one’s by 𝟑𝐁𝐥𝐮𝐞𝟏𝐁𝐫𝐨𝐰𝐧, basically my best online “teacher” back when I was studying NLP at uni. Still one of the best free resources for understanding deep learning fundamentals.

(YouTube channel has the same name.)

Also, if you want a hands-on intro to how modern AI really works in practice, 𝗢𝘂𝘁𝘀𝗸𝗶𝗹𝗹 is running a live program this weekend.

Live session | Free to join
🔗https://lnkd.in/gdKQFYBk
__________

For more on AI and learning materials, plz check my previous posts. 

I share my journey here. Join me and let's grow together. Alex Wang

Open source is winning again.
𝗞𝗶𝗹𝗼 𝗖𝗼𝗱𝗲, built by the co-founder of GitLab, with a real engineering mind behind it.

Explore here: https://lnkd.in/gASAa6zq

Already the  #𝟭 𝗚𝗿𝗼𝗸 𝟭 𝗰𝗼𝗻𝘀𝘂𝗺𝗲𝗿 -
and growing fast: 20+ updates in the last 30 days, 400K+ downloads in under 6 months.

• 𝗕𝘂𝗶𝗹𝘁 𝗳𝗼𝗿 𝘁𝗲𝗮𝗺𝘀
 Shared model access, persistent memory, consistent context

• 𝗧𝗿𝗮𝗻𝘀𝗽𝗮𝗿𝗲𝗻𝘁 𝘁𝗼 𝘁𝗵𝗲 𝗰𝗼𝗿𝗲
 Model-agnostic and fully self-hostable

Created in partnership with Kilo Code, a tool I’ve been meaning to explore.

Connect once. Use 9,000+ tools right in ChatGPT Atlas.

Every new AI tool means another integration to build and maintain.

With Apify MCP connected to OpenAI (ChatGPT Atlas), you get instant access to 8,000+ web scrapers, APIs, automations, and agents through a single connector with no endless integrations.

---- What you can do

🔹 𝗢𝗻𝗲 𝗰𝗼𝗻𝗻𝗲𝗰𝘁𝗼𝗿 → thousands of actions (scrapers, APIs, automations, agents)

🔹 𝗪𝗼𝗿𝗸 𝘄𝗵𝗲𝗿𝗲 𝘆𝗼𝘂 𝗮𝗿𝗲 → trigger tools from any tab in Atlas

🔹 𝗖𝗵𝗮𝗶𝗻 𝗰𝗼𝗺𝗽𝗹𝗲𝘅 𝘄𝗼𝗿𝗸𝗳𝗹𝗼𝘄𝘀 → mix research, enrichment, search, and more

🔹 𝗥𝗲𝘀𝘂𝗹𝘁𝘀 𝗳𝗹𝗼𝘄 𝗯𝗮𝗰𝗸 𝗶𝗻𝘀𝘁𝗮𝗻𝘁𝗹𝘆 → into your doc, sheet, or dashboard

---- Real wins

• Research ops: scrape, clean, summarize, and drop into Sheets
• GTM & RevOps: on-demand company/contact enrichment
• Travel & e-com: live inventory/price checks with links to act
• Competitive watch: monitor pages and pipe insights to your workspace

One setup. Thousands of actions. No integration backlog.

𝗖𝗿𝗲𝗮𝘁𝗲 𝘆𝗼𝘂𝗿 𝗼𝘄𝗻 𝗠𝗖𝗣 𝘀𝗲𝗿𝘃𝗲𝗿 → https://lnkd.in/gkGibMun

#mcp #generativeai #llms #artificialintelligence

No more prompts, just command - Rocket made it happen.

If you’ve ever built with AI, you know the cycle:
write a prompt → AI misinterprets → fix → retry → clarify → hope nothing breaks.

Half the time is 'babysitting' the model.

Rocket just launched something flips that:
Precision Mode — 100+ structured commands for building full apps.

I tested it this week. Here’s how it plays out:

𝟭) 𝗔𝗱𝗱 𝗽𝗮𝘆𝗺𝗲𝗻𝘁𝘀 (𝗦𝘁𝗿𝗶𝗽𝗲)

Goal: add a one-time checkout
Command: /Add Stripe Payments
Result: Rocket generated the full payment flow instantly, checkout screen, integration, test-ready.

𝟮) 𝗙𝗶𝘅 𝗺𝗼𝗯𝗶𝗹𝗲 𝗹𝗮𝘆𝗼𝘂𝘁

Goal: clean up spacing and responsiveness
Command: /Fix Layout Issues
Result: Everything snapped into place — aligned, device-responsive, production-ready.

𝟯) 𝗔𝗱𝗱 𝗮𝗻 𝗔𝗜 𝗰𝗵𝗮𝘁𝗯𝗼𝘁

Goal: embed an assistant powered by OpenAI or Anthropic
Command: /Add AI Chatbot
Result: Fully functional chatbot added, configured, and embedded, no manual setup.

𝟰) 𝗠𝗮𝗸𝗲 𝘁𝗵𝗲 𝗮𝗽𝗽 𝗺𝘂𝗹𝘁𝗶𝗹𝗶𝗻𝗴𝘂𝗮𝗹

Goal: prepare the entire app for multiple languages
Command: /Add Multiple Languages
Result: Rocket structured the project for full localization across screens, instantly.

It’s fast, predictable, and removes the ambiguity.

📍Test it yourself for free here https://lnkd.in/gPc_rRWu

𝗦𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲𝗱 𝗰𝗼𝗺𝗺𝗮𝗻𝗱𝘀 > 𝗴𝘂𝗲𝘀𝘀-𝗯𝗮𝘀𝗲𝗱 𝗽𝗿𝗼𝗺𝗽𝘁𝗶𝗻𝗴
This might be what the next chapter of app development looks like.

Stop making PowerPoints - your next deck can start with a Word doc!

Plus AI ‘s latest AI feature takes your existing docs and turns them into well-formatted, professional presentations - automatically.

You can:
- Select from dozens of professionally-designed templates
- Create advanced slide types with native charts, tables, and case studies
- Translate existing presentations into any language
- Best part: Plus AI creates actual PowerPoint files

It’s easy to get started, and you can try it for FREE here: https://lnkd.in/gj54VYuM 

#aitools #productivity #bussiness #generativeai

SOC 2 is one of the most underestimated revenue blockers for teams.

It hits every company:
If your SOC 2 isn’t ready, your product demo doesn’t matter — the deal stalls right there.

If you want to clear that bottleneck fast, Delve is running a Thanksgiving offer this week:

$𝟭,𝟱𝟬𝟬 𝗼𝗳𝗳 𝗦𝗢𝗖 𝟮 𝗰𝗼𝗺𝗽𝗹𝗶𝗮𝗻𝗰𝗲 + 𝗮 $𝟱𝟬𝟬 𝗩𝗶𝘀𝗮 𝗴𝗶𝗳𝘁 𝗰𝗮𝗿𝗱

💡How they help:

SOC 2 is normally a time sink.
Evidence buried across tools, endless spreadsheets, manual risk checks… everything slows down.

Delve’s AI agents take over the grunt work — pulling evidence, mapping controls, running checks, prepping your audit — so teams can go from “blocked” to “compliant” in days.

🔥Some recent wins:

▪️Lovable → SOC 2 in 20 hours

▪️WisprFlow → Closed Mercury + Superhuman in 2 weeks

▪️11x → $1.2M ARR unlocked

If SOC 2 is slowing your pipeline, this is a good week to get it done — quickly, cleanly, and without losing your sanity.

📍Link in the comments.
Code: 𝗧𝗛𝗔𝗡𝗞𝗦𝗚𝗜𝗩𝗜𝗡𝗚, offer ends 11/27 at 11:59 PM PST.

This movie-quality video started with just six images.
Everything in between, the motion, lighting, and movements…was generated by AI.

Tried Freepik recently and it was smoother than expected.

The flow is simple but surprisingly flexible:
 → Pick key moments as stills
 → Let the AI build transitions — movement, mood, pacing
 → Fine-tune with tools like 𝗘𝘅𝗽𝗮𝗻𝗱, 𝗥𝗲𝘁𝗼𝘂𝗰𝗵, and 𝗨𝗽𝘀𝗰𝗮𝗹𝗲

What stood out most: you still have a high level of creative control.
It’s not a black box: you guide the direction, and the AI fills in the story.

This rabbit scene went from a simple concept to a full cinematic mini-sequence in minutes.

If you’re building visuals for content, ads, or quick storytelling, this kind of workflow really speeds things up - without sacrificing quality or vision.

📍Create smarter with Freepik https://lnkd.in/g-73QFgz

#aitools #productivity #generativeai

Yugen.ai

Developed and maintained a recommendation system, optimizing models for diverse countries and customer segments using AWS Sagemaker.
Applied boosting algorithms like XGBoost, LightGBM and Decision Trees to optimize model performance.
Delivered valuable insights from Big Data through proficient Data Analysis, SQL and BigQuery.
Contributed to creating the deep learning architecture for the AutoML solution on Sagemaker.

Lynk Global

Adeptly analyzed text data, conducted topic modelling, and applied  Natural Language Processing (NLP) techniques to cluster and label documents, enhancing data insights.
Deployed, tested, and maintained Natural Language Processing (NLP) solutions through a CI/CD pipeline, ensuring seamless and efficient operations and deployment.
Led the development and implementation of data handling and cleaning procedures for personal data from diverse sources, systematically removing duplicates and ensuring secure data storage.
Applied Natural Language Processing (NLP) techniques to enhance data insights. Conducted Data Analysis, specializing in Keyword Extraction and Text Similarity assessments.

Data Science Engineer

Activa Inc

Managed an 8-member Data Scientist team, leading the AutoML project.
Collaborated cross-functionally with front-end development, design, digital marketing, and data engineering teams to realize product objectives.

Data Scientist (Team Lead)

Spearheaded AutoML solution development for Time Series data.
Directed end-to-end AutoML implementation, encompassing pre-processing, feature engineering, algorithm selection, optimization, and deployment.
Pioneered Genetic Algorithm utilization for model optimization.
Drove computer vision projects, including Object Detection for machinery defects and electrical components.

Associate Data Scientist

SmartDocs Inc

Held the position of overseeing all data science-related initiatives, leading the conception and execution of data-driven solutions.
Implemented image classification using YOLO, TensorFlow and Keras, specifically for industrial machines.
Conducted text similarity assessments using fuzzy logic.
Created chatbots using RASA and Microsoft Chatbot services.

Data Scientist | Machine Learning | Deep Learning | Natural Language Processing

Data Scientist with 3+ years of experience in NLP, deep learning, and machine learning. Passionate about using data to solve real-world problems. Seeking a data-driven, flexible, and innovative company in Germany, preferably Berlin or Cologne, where I can implement my skills and make a positive impact.

I've worked on various projects in the past three years, including leading a team at BotIQ in Hyderabad, India. My favourite field is Natural Language Processing (NLP), which I worked on at LYNK Global. I also have experience in Computer Vision, Machine Learning, and Data Analytics.

I recently got my MBA to expand my business knowledge and better understand how to solve business problems. I'm looking for a place to use my skills and experience as a data scientist to make a positive impact. I'm a team player, always seeking ways to learn and grow.

Area of Interest - Natural Language Processing (NLP), Deep Learning, Machine Learning
Domains of Interest - Finance, Cryptocurrency, Technology, Artificial Intelligence, Big Data, Chatbots
Programming Languages: Python, R

Frameworks or Libraries:
Pandas, Tableau, MS Excel (Data Analysis)
TensorFlow, PyTorch, Keras (Deep Learning)
Scikit-Learn, Statsmodels (Machine Learning)
Spacy, Gensim (Natural Language Processing)
Matplotlib, Seaborn, Tableau (Data Visualization)

taplio

The Battle

Alex Wang

Kavita Aherwar

Followers

Posts

Engagements

Likes

Performance ScoreAlex Wang's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Alex Wang

Kavita Aherwar

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);}Alex Wang's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreAlex Wang's LinkedIn Performance