GoWit-Hungary

We are a Software Development Company with a focus on bridging our European based services with our Vietnamese team. We specialise in building tailored teams.

[GoWit-Hungary @GowitHungary]

We're happy to share that our CEO @SylvainLiege obtained a new certification:
AWS Certified AI Practitioner certificate from Amazon Web Services (AWS)!

--> We provide AI services to our customers.<-- we'd be happy to answer your questions and contribute, with our teams, to the successes of your projects.

@SylvainLiege is also writing papers on AI to make AI understandable to everyone: "The Hunt for Intelligence".

[GoWit-Hungary @GowitHungary]

Did you know Vietnam is producing 50,000 IT engineers every year, building a workforce of over 500,000 professionals? This isn’t just a numbers game, it’s the result of a deliberate government push to make Vietnam a global tech hub, with investments in STEM education. The result? A young, dynamic talent pool (over half the population is under 35!) driving innovation in software development, AI, and digital transformation.

What makes Vietnam stand out? It’s not just the scale—it’s the value. Businesses are discovering that Vietnam offers high-quality IT talent at competitive costs, without compromising on delivery.
At Liemur and GoWitHungary, we’ve seen this firsthand, blending Vietnam’s tech energy with our UK and Hungary operations to solve real-world challenges for clients.

Is Vietnam on your radar for IT partnerships? Share your thoughts in the comments or DM me. I’d love to hear how you’re navigating the global tech landscape!

#VietnamTech #ITInnovation #TechTalent #DigitalTransformation

[GoWit-Hungary @GowitHungary]

@minchoi When you write "Never pay for online courses again", it is because AI has stolen the know how to some humans who cared doing it. How is this a victory of any sort?

I am not against AI at all, but I am surprised about what people are celebrating about, sometimes...

[GoWit-Hungary @GowitHungary]

Choosing the right off-the-shelf AI model starts with knowing the precise business needs, not just throwing tech at the wall to see what sticks !

Let’s talk about an inconvenient fact in AI: not every shiny off-the-shelf model is your golden ticket. You can grab a pre-trained model faster than you can say “neural network,” but if you don’t assess properly your business needs first, you’re registering for a not-so-fun AI ride.

Here’s the deal: AI models aren’t one-size-fits-all. They’re built with specific data, for specific tasks, with specific biases (topic of my coming AI paper #10!). If your business problem doesn’t match the model’s “experience,” you’re setting yourself up for a very expensive trip. It’s like asking a Formula 1 car to haul furniture—sure, it’s fast, but your sofa’s still on the sidewalk.

Step 1: Know the business like the back of your hand. What’s the goal? Who’s the user? What’s the data? What’s the risk if it goes wrong? If you skip this, you’re basically handing your wallet to the AI gods and saying, “Surprise me!” (Nah…you won’t like it!)
Example: Imagine you run a small e-commerce shop selling artisanal soaps. You want an AI to recommend products to customers. Sounds simple, right? So, you grab a fancy off-the-shelf recommendation model—let’s call it “Hopeforsales 3000”—trained on millions of Netflix viewing habits. It’s great at suggesting the next binge-worthy series, but soaps? Soap operas, maybe… It starts recommending lavender soap to someone who just bought… lavender soap. Why? Because it’s obsessed with patterns like “if they watched one crime drama, suggest another,” not “if they bought a soap, suggest a complementary scent.” Your customers get annoyed, sales tank, and you’re left wondering why your AI is playing matchmaker with the wrong crowd.
Now, rewind. Assess your business first: you’ve got a niche product, a small but loyal customer base, and limited data—maybe purchase histories and some reviews. You need a model that thrives on sparse data, understands product pairings (like soaps and lotions), and prioritizes customer delight over volume. A lightweight collaborative filtering model, pre-trained on e-commerce data (think Etsy, not Netflix), would’ve done marvel. Instead of “Hopeforsales 3000,” you’d have “MySalesExpert 3000”, happy customers and a fatter wallet.
The moral? AI isn’t magic, it’s math, and math needs the right context. In my papers (#7, #9), we saw how neural networks are just functions mimicking patterns. If the pattern doesn’t fit your business, you’re going nowhere good. So, before you grab that off-the-shelf model, do your homework. Assess your business, define your goal, and pick the model that’s been “trained” for your fight.

What’s your take? Ever picked the wrong model and learned the hard way? Let’s chat in the comments!

If you fancy more in-depth knowledge, here are my AI papers: https://t.co/kHrcgiXIup

hashtag#AI hashtag#BusinessStrategy hashtag#MachineLearning

[GoWit-Hungary @GowitHungary]

Want to slash costs on your AI project without skimping on results? Enter RAG—Retrieval-Augmented Generation, the clever trick that’s like giving your AI a library card instead of a PhD tuition bill! It’s cheaper and faster.

Here’s the deal: training massive AI models from scratch is like building a rocket to the moon—expensive, time-consuming, and overkill if you just need to pop to the corner shop. RAG says, “Hold my coffee.” Instead of retraining a model every time you need new info, RAG grabs relevant data from your existing documents, databases, or knowledge bases and feeds it to the AI to generate spot-on answers. Think of it as a super-smart assistant who knows where to look for answers rather than memorizing the entire encyclopedia.

Why does this save you money?

* No retraining tax: Updating a model’s knowledge usually means burning cash on compute power and data scientists tweaking parameters (all those w and b we talked about in my Neural Network papers!). RAG sidesteps this by pulling fresh info on the fly.

* Leaner infrastructure: You don’t need a monstrous model to store all the world’s knowledge. A smaller model plus a good retrieval system does the job, cutting down on GPU bills (AWS, we still like you but…).

* Faster deployment: RAG lets you plug in new data sources without starting from scratch. Got a new product manual or customer FAQ? Toss it in the retrieval system, and your AI’s ready to roll.

Picture this: you’re building a customer support AI. Without RAG, you’d train it on every possible question, costing a fortune and still missing edge cases. With RAG, you point it to your support docs, and it pulls the right answers in real-time—accurate, up-to-date, and cheap. It’s like hiring a librarian who’s also a poet, not a team of scholars rewriting the library.

Now, is this “intelligence”? Nah, it’s just humans being clever with data and math (as usual). RAG’s genius is in the design—combining retrieval and generation to keep your wallet happy while delivering results that feel like magic.

What’s your take? Have you tried RAG in your AI projects, or are you still burning cash the old-fashioned way? Let’s chat in the comments!

hashtag#AI hashtag#RAG hashtag#CostSaving hashtag#MachineLearning

[GoWit-Hungary @GowitHungary]

I’ve attended the Hong Kong Electronics fair this last week-end. The scale of the event was colossal, of course, with almost 2000 participants to present products.
There was a huge number of already-seen products like smart scales, power banks, watches, daily products like clippers, smart kitchen bins or massagers, and you could not miss the audio video devices present.

The star of the fair was the concept of smart glasses and smart ear pods. Both were supposed to do amazing things, the most important being live translation.
To be honest, I would have paid a lot for a Chinese to English live translator. Unfortunately, every time I asked to try the devices, there was a reason not to. The features were not always clear either, for instance, I was told that one pair would display information but when I asked if it would be displayed in the glasses themselves, the answer was fuzzy and it could very well be that the displaying was on your phone.
Although it was not for testing, we could have a taste of our not so far future.
My favorite products I have seen are:
•⁠ ⁠A neck massager that was a real delight, especially after hours of standing in the crowd. It was heating at perfect temperature and doing a very nice job at pressing at the right places. They had the feet massager going with it but I could not try it, people were queuing to try. I did not have the stamina to queue for that.
•⁠ ⁠A kitchen smart bin that converts all your organic rubbish into powder, including bones, but also plays the role of a music player and an ambient light. The lady owning the product was very proud of it and very friendly.
•⁠ ⁠A crazy cheap party music player that was incredibly light in weight, good in sound and cheap in price. The thing was costing $19.90 with a sound without any distortion, even very loud. Amazing!
•⁠ ⁠A special teacup form Hanking that keeps your tea at the perfect temperature. As a tea-guy, this one was very appealing. Maple Mao was proud of her products and rightly so.

If you want to be overwhelmed by electronics products, it’s the place to go to!

[GoWit-Hungary @GowitHungary]

# 7 - AI Architecture: Neural Network Design
https://t.co/YpdEhx5W4d

#Summary: AI Neural networks mimic the neural network of the brain. Once the technical architecture has been built, how does each component work? We present the various mathematical component in action. This is paper #7 in the series.

We recommend reading the previous articles in the series for an easier understanding. https://t.co/kHrcgiXaER

#Keywords: AI; Neural network; Machine Learning; Deep learning; Neuron; Synapses; Layers; Embedding; Vectorization; Activation Function; Feature Selection; Encoding; Forward Propagation

Author: Sylvain LIÈGE

Note: This Paper was *NOT written by AI*.

[GoWit-Hungary @GowitHungary]

# 7 - Artificial Intelligence Architecture: Neural Network Design
https://t.co/YpdEhx6tTL

#Summary: AI Neural networks mimic the neural network of the brain. Once the technical architecture has been built, how does each component work? We present the various mathematical component in action.

This is paper #7 in the series. We recommend reading the previous articles in the series for an easier understanding.
https://t.co/kHrcgiXIup

#Keywords: AI; Neural network; Machine Learning; Deep learning; Neuron; Synapses; Layers; Embedding; Vectorization; Activation Function; Feature Selection; Encoding; Forward Propagation
Author: Sylvain LIÈGE

Note: This Paper was *NOT written by AI*.

[GoWit-Hungary @GowitHungary]

White Paper in the Series: AI: The Hunt for Intelligence!

#08 - AI: Neural Network – Forward Propagation https://t.co/oPhV47N9sz

Summary: AI Neural networks mimic the neural network of the brain. In this paper we present what is happening inside a digital neural network from data entry to result. We study the various mathematical steps in their simplest format to allow global understanding of the inside mechanisms. The end-to-end process is called Forward Propagation.

We recommend reading the 7 previous articles in the series for an easier understanding. https://t.co/kHrcgiXIup

Keywords: AI; Neural network; Machine Learning; Deep learning; Neuron; Synapses; Layers; Embedding; Vectorization; Activation Function; Forward propagation

Author: Sylvain LIÈGE

Note: This Paper was NOT written by AI, although AI might be used for research purposes.

[GoWit-Hungary @GowitHungary]

White Paper in the Series: AI: The Hunt for Intelligence!

#08 - AI: Neural Network – Forward Propagation https://t.co/oPhV47MBD1

Summary: AI Neural networks mimic the neural network of the brain. In this paper we present what is happening inside a digital neural network from data entry to result. We study the various mathematical steps in their simplest format to allow global understanding of the inside mechanisms. The end-to-end process is called Forward Propagation.

We recommend reading the 7 previous articles in the series for an easier understanding. https://t.co/kHrcgiXaER

Keywords: AI; Neural network; Machine Learning; Deep learning; Neuron; Synapses; Layers; Embedding; Vectorization; Activation Function; Forward propagation

Author: Sylvain LIÈGE

Note: This Paper was NOT written by AI, although AI might be used for research purposes.

[GoWit-Hungary @GowitHungary]

White Paper in the Series: AI: The Hunt for Intelligence!

#08 - AI: Neural Network – Forward Propagation
https://t.co/oPhV47MBD1

Summary: AI Neural networks mimic the neural network of the brain. In this paper we present what is happening inside a digital neural network from data entry to result. We study the various mathematical steps in their simplest format to allow global understanding of the inside mechanisms. The end-to-end process is called Forward Propagation.

We recommend reading the 7 previous articles in the series for an easier understanding.
https://t.co/kHrcgiXaER

Keywords: AI; Neural network; Machine Learning; Deep learning; Neuron; Synapses; Layers; Embedding; Vectorization; Activation Function; Forward propagation

Author: Sylvain LIÈGE

Note: This Paper was NOT written by AI, although AI might be used for research purposes.

[GoWit-Hungary @GowitHungary]

White Paper: Artificial Intelligence: Digital Neural Network Architecture
https://t.co/otC6VRKR1d

Topic: AI Neural Network : A digital architecture to mimic the brain Summary: AI Neural networks mimic the neural network of the brain. But how do build a digital neural network? What is its architecture? We present the basic component of such technical solution.

No mathematical or scientific background is needed to read this paper.

Keywords: AI; Neural network; Machine Learning; Deep learning; Neuron; Synapses; Layers

Note: This Paper was NOT written by AI, although AI might be used for research purposes.

[GoWit-Hungary @GowitHungary]

White Paper: Artificial Intelligence: Digital Neural Network Architecture
https://t.co/otC6VRKR1d

Topic: AI Neural Network : A digital architecture to mimic the brain Summary: AI Neural networks mimic the neural network of the brain. But how do build a digital neural network? What is its architecture? We present the basic component of such technical solution.

Keywords: AI; Neural network; Machine Learning; Deep learning; Neuron; Synapses; Layers Note:

This Paper was NOT written by AI, although AI might be used for research purposes.

[GoWit-Hungary @GowitHungary]

White Paper: Artificial Intelligence: Digital Neural Network Architecture
https://t.co/otC6VRKR1d

Topic: AI Neural Network : A digital architecture to mimic the brain

Summary: AI Neural networks mimic the neural network of the brain. But how do build a digital neural network? What is its architecture? We present the basic component of such technical solution.

No mathematical or scientific background is needed to read this paper.

Keywords: AI; Neural network; Machine Learning; Deep learning; Neuron; Synapses; Layers

Note: This Paper was NOT written by AI, although AI might be used for research purposes.

[GoWit-Hungary @GowitHungary]

White Paper:
Basis of  AI : Digital Neural Network derived from biology

https://t.co/iIBFdowlLX

Summary: AI Neural networks mimic the neural network of the brain. What are the principles driving a neural network? How did we look at biology to create the most powerful “machines” ever created?

Keywords: AI; Neural network; Machine Learning; Deep learning; Neuron; Synapses; Layers

Note: This Paper was NOT written by AI.

[GoWit-Hungary @GowitHungary]

White Paper:
Basis of  AI : Digital Neural Network derived from biology
https://t.co/iIBFdowlLX
Summary: AI Neural networks mimic the neural network of the brain. What are the principles driving a neural network? How did we look at biology to create the most powerful “machines” ever created?

Keywords: AI; Neural network; Machine Learning; Deep learning; Neuron; Synapses; Layers Note:

This Paper was NOT written by AI, although AI might be used for research purposes.

[GoWit-Hungary @GowitHungary]

White Paper: Basis of  AI : Digital Neural Network derived from biology
https://t.co/iIBFdowlLX
Summary: AI Neural networks mimic the neural network of the brain. What are the principles driving a neural network? How did we look at biology to create the most powerful “machines” ever created?

Keywords: AI; Neural network; Machine Learning; Deep learning; Neuron; Synapses; Layers

Note: This Paper was NOT written by AI, although AI might be used for research purposes.

[GoWit-Hungary @GowitHungary]

Topic: Mathematics used in AI : Differential Calculus

Short paper on differential calculus for AI !
https://t.co/LyH1b3XP8N
Summary: Differential Calculus is the corner stone of training a ML model. It is the key to predict the future correctly, based on the past.

Note: This short paper does NOT require mathematical background.

Keywords: AI; Differential Calculus; Machine Learning; Loss Function

[GoWit-Hungary @GowitHungary]

Artificial Intelligence & Mathematics: Algebra

Short paper on Algebra for AI !
https://t.co/vCZlUnKAiy

Algebra is an essential mathematical enhancer for AI. It allows to represent the real world in mathematical structures that can be easily manipulated by the computer.

This short paper does not require mathematical background.

[GoWit-Hungary @GowitHungary]

Topic: Mathematics used in AI : Differential Calculus

Summary: Differential Calculus is the corner stone of training a ML model. It is the key to predict the future correctly, based on the past.

Note: This short paper does NOT require mathematical background.

Keywords: AI; Differential Calculus; Machine Learning; Loss Function

https://t.co/LyH1b3XP8N