Data Science AI

Leveraging a diverse skill set to help businesses transform their operations, enhance productivity, and drive growth through the power of data and AI. I count on my experience in business strategy, project leadership, data science, and enterprise architecture to help companies. I solved business challenges and enabled data-driven decision-making in various industries like IT, software, CPG, transport and logistics, professional services, fintech, and outsourcing.
★ Help companies unlock their growth potential through data science and AI.
★ Translate complex technical concepts into actionable insights for business leaders.
★ Work with diverse businesses to design and build strategic data-driven decision-making systems.
★ Designed and implemented an automated end-to-end decision-making system, integrating data across multiple sources into an Azure MySQL data warehouse and developing a Shiny web application on AWS EC2 with machine learning features for enhanced C-level decision-making.

Data Science Consultant

Blankfactor

I used a mixture of leadership, project/process management, and technical skills to design robust data science applications. I worked with business clients or internal stakeholders, translating business needs into data-driven decision-making solutions.
★ Worked on employee turnover analysis that revealed an opportunity for $640k per year in productivity gains. Deployed a web app with R & Python (employee grouping, trained classification models with H2O and R tidy models framework).
★ Built internal structure and processes for effective team collaboration. Lead the internal data practice - managing a team of data practitioners, and administering the main tools like Jira/Confluence, AWS, Snowflake, and Google Workspace. 
★ Designed a robust cloud-based architecture leveraging Snowflake, integrating various data sources that enhanced the data science and AI capabilities of the company.

Senior Data Scientist / Data Officer

iCOVER

I leveraged industry best practices to define the transformational strategy of the company on a global level. I collaborated with all teams to ensure the completion of various projects related to IT, HR, Product, and Delivery. I first led a team of Product Managers who developed new products related to background screening. I led a team focused on improving the company's data and project maturity. I also worked on several data engineering and data science activities. Accountable for a local site of 100+ people distributed in different teams.

★ Data Management: deployed global data science capabilities with python data engineering, MySQL, and Tableau Online visualization. Worked on predicting customer churn with Spark, and deployed a testing disaster response classifier web app on Heroku cloud. Used clustering to group clients and explore revenue-generating products. Worked with MS Azure cloud capabilities to streamline the teamwork. 
★ Business Process Management: designed, and deployed a document and process infrastructure based on Google Workspace. Established a quality management system with process governance according to CMMI Model. Increased organizational maturity from level 1 to level 3.
★ Project Portfolio Management - designed, and deployed a hybrid model for project delivery within Jira Software/Confluence that enabled complete track of scope, time, and cost. Deployed Jira Service Desk in IT and operational teams that covered 30% of all company population.
★ Leadership: coordinated the efforts of renting a new building for ~100 employees in Sofia. Organized a project team and checked more than 16 potential locations.

Projects and Data Lead

Cargill

I developed a business process management strategy to achieve organizational goals while leveraging IT service management best practices. I collaborated with cross-functional teams to ensure the completion of various projects related to financial, HR, IT, continuous improvement, and change management. I led a team of eight Project Managers who were responsible for migration, improvement, and deployment projects/programs. Used Lean Six Sigma tools to run statistical experiments on a broad level of data related to finance, logistics, HR, and IT.

★ Designed and implemented a Lean Six Sigma certification program for a local site of 1100 people and managed 12 Yellow Belt and 20 Green Belt candidates to improve various processes related to Finance, Procurement and Logistics. 
★ Drove improvement by identifying and mitigating the potential risk of non-compliance as well as streamlined standard operating procedures by implementing a regional business process management framework.
★ Designed Global Project and Portfolio Management Framework based on Clarizen tool that increased portfolio control and visibility.

Project Portfolio Lead

I updated executive management with respect to business progress. I headed a highly skilled team of international experts and conducted a cost-benefit analysis for multiple programs. I defined project road maps, schedules, and work breakdown structures to achieve desired goals. I delivered a full range of quality, BCP, lean six sigma, and project management consultancy services to promote E2E collaboration between the EMEA teams.

★ Spearheaded, completed, and delivered a number of major projects within time and budgetary constraints.
★ Identified cost reduction opportunities for $500k in the commercial delivery process and resolved the root cause of issues in link with key HP client to renew $600k contract deal.
★ Created BCP partner assessment framework to mitigate risks and increase HP customer support resiliency.

Quality Program Manager

I devised a business strategy for conflicts management, performance enhancement, training, and development. I administered reporting team that was engaged in delivering business analysis and SharePoint services to improve business performance. I contributed to the provision of comprehensive project support in the GSD Supply chain as a material buyer.

★ Minimized inventory cost by 50% through optimization of an efficient reactive sourcing spare parts process.
★ Decreased report generation time by 30% through automation of key performance indicator delivery process.
★ Ensured smooth running of supply chain operations by developing project strategies for resource distribution.

Business Analyst

Familex Ltd

I managed logistics projects, including supply chain optimization, new payment terms, and international relations. I coordinated with complex vendors to improve trade processes. I promoted international business and expanded the company’s portfolio during several international fairs.

★ Developed an effective logistic model for stable supply deliveries to decrease the average time to deliver by 38%.
★ Optimized operational performance by providing detailed training to meet the requirements of the trading department.
★ Negotiated contracts with key suppliers to reduce purchase price by 20% and save around $300k annually.
★ Designed a third-party delivery inspection model and letter of credit bank payment, slashing the non-confirming deliveries from 30% to 0.

Manager, International Trade

Data Scientist | Consultant | Lean Six Sigma MBB

From Lean Six Sigma projects to data science solutions, I've turned complex business problems into strategic opportunities. My key to success? Bridging the gap between business and technology. With expertise in project leadership, data science, and business strategy, I help companies unlock their potential using data-driven decision-making.

My journey has led me through the corridors of Fortune 50 giants like Cargill and HP, where I led various teams, and implemented tech solutions for business success. Then I left the big corporate world and I witnessed first-hand the challenges of the small-medium sized businesses. My work in time series forecasting, customer segmentation, and attrition prediction has informed strategic direction and uncovered untapped opportunities.

Today, as a consultant, I'm passionate about extending this impact beyond corporate boundaries. I leverage my diverse skill set to help businesses transform their operations, enhance productivity, and drive growth through the power of data and AI.

If you're ready to leverage data science and AI for your business, I'm here to catalyze that journey. Let's partner to navigate the data-driven future together.

🤝 Consultation Enquiries: https://data-science-ai.com/applynow/
 🐦 Twitter: https://twitter.com/p_rabadzhiyski

RPA - the next big thing! 

A title back in 2017... 

What is RPA (Robotics Process Automation) and is it AI? Let's dive in.

I still remember a conference back in June 2017. 

It was organized by one of the leading consulting companies in the world. It was all about RPA (Robotics Process Automation), and its incredible potential to solve business problems.

I was so impressed by the idea of having a bot to do the boring things we do as humans during our work. Fascinating. 

I smile today, and think: That hype died long ago but is RPA part of AI?

Robotic Process Automation is not AI.

While RPA is used to work in conjunction with people by automating repetitive tasks (attended automation), AI is viewed as a form of technology to replace human labor and automate end-to-end (unattended automation). 

RPA uses structured inputs and logic, while AI uses unstructured inputs and develops its own logic. 

In essence, RPA uses "bots" to replicate human interaction with desktop applications by using rule-based automation.

Bots/Robots are a virtual workforce governed by the business operations teams.

RPA mimics employee's execution of repetitive tasks via existing user interfaces.

It is best suited for relatively simple and linear tasks, which are highly repetitive. 

AI can complement a Robotics Process Automation. 

For example, AI can bring decision-making and analytical capabilities which in turn can improve the performance of a Robotics Process Automation system.

To sum up

Robotics Process Automation is not AI. It's a way to automate simple and linear tasks with the help of virtual assistants (bots). The bots can work 24/7 but their performance is governed by humans. 

We can probably say that RPA is process-driven, while AI is data-driven.

🤖 🤖 🤖 

#rpa #ai #processimprovement

Plamen Rabadzhiyski

I Can Help You Make Data-Driven Decisions 📊 Data Science & AI Consultant 💡 Let's Solve Your Toughest Challenges Together! 🔍

Lost in translation - machine learning vs statistical learning. 

Are they the same? 

It took me a lot of time, and research to find the answer for me. 

Both terms are kind of interchangeable now but there is a difference. Let me explain.

Statistical learning: 

Emerged from statistics, focusing on understanding data through models grounded in statistical theory. It aims at inference, seeking to explain relationships within the data. 

Machine learning:

Originated more from computer science, with a focus on enabling computers to learn from and make decisions or predictions based on data. It often emphasizes predictive accuracy and the ability to handle large volumes of data, even at the cost of model interpretability. 

Over time, as data has become more abundant and computing power has increased, the fields have converged in many respects. 

Machine learning techniques now often incorporate statistical principles, while statistical learning increasingly uses computational methods to deal with complex data. 

This convergence has led to some using the terms interchangeably, especially in practical applications where the distinction between developing a predictive model and understanding data structures may blur. 

Today, "machine learning" has become the more popular term, especially in the tech industry and among practitioners focusing on developing AI applications. 

This popularity reflects the field's emphasis on practical/predictive applications and its association with the broader AI domain. 

#machinelearning #statistics #data

What is Deep Learning and how do we use it today?

I've been struggling in the beginning of my career to make sense of all that terms. 

I finally realized that AI is the broader term. 

It goes that way: AI->ML->DL->who knows what?

Machine Learning is a subset of AI. Deep Learning is a subset of Machine Learning.

Most of the AI-related inventions were inspired by how the human brain works. 

Deep Learning is a great example. It mimics the neural networks of the brain. 

Today computers use artificial neural networks to process and analyze information.

Neural networks consist of nodes layered together. When a network is composed of three or more layers it is said to be "deep". Hello, deep learning!

With time, neural networks evolved, and several enhanced "versions" were created. 

In fact, the Generative AI explosion in recent years was due to the new approaches in building neural networks (e.g. Transformers). 

Deep learning is the backbone of some of the most complex systems today - from computer vision, through self-driving cars, and large language models. 

However, this fascinating technology has its downsides - it's a complete black box. 

No one can tell why a deep learning system generated its final output. That problem became a whole new research field. 

On top of that, deep learning needs data (a lot), and fast compute power (e.g. GPUs hardware).

To sum up

Deep learning is a subset of machine learning that uses artificial neural networks to process and analyze information. It's greedy for data, and it's the backbone of the modern AI systems.

 #deeplearning #ai #machinelearning

Is machine learning AI? 

A logical question or just a misconception? 

Let's find out together.

I got confused at the beginning of my journey with data science. I will try to bring some clarity based on my research. 

AI (Artificial Intelligence)

It's a broad field, which refers to using technology to build machines/computers that can mimic human intelligence.

For example, being able to see, understand and respond to spoken/written language, analyzing data, making recommendations, and more. 

ML (Machine Learning)

It's a subset of AI, that automatically enables a system/machine to learn and improve from experience.

Instead of explicit programming, it uses algorithms to analyze data, learn from the insights, and make informed decisions. The more data you have the better.

To sum up

AI is the broader concept - enabling a machine to sense, reason, or adapt like a human

ML is an application of AI - allowing machines to extract knowledge from data and automatically learn from it.

So, AI is a broad field, ML is a narrow application/specific approach/ to learn from data.

I hope it makes sense. However, what about deep learning?

That will come in other posts :) 

# datascience #machinelearning #ai

An intro to a fascinating world of technology and human ingenuity.

2020 was when I first heard about NLP (Natural Language Processing).

It wasn't so interesting for me back then, but I got hooked two years later with the explosion of ChatGPT.

I have always struggled to understand how we got where we are, but at some point it made sense.

ChatGPT-like tools walked a long road and the advancements in Natural Language Processing made all that possible.
Let's dive in.

NLP is a subset of AI that helps computers understand, interpret, and utilize human language.

The main approach is to break the language into shorter pieces called tokens, and then try to understand the relationship of those tokens.

It all started back in the 1950s when AI was an even more popular word than today.

With time, new approaches were introduced with the sole objective of making computers interpret human language.

The efforts went through ups and downs until the beginning of the 21st century when neural networks started taking over the field.

2001 was the first time when an artificial neural network was used for understanding text.
With a feed-forward neural network that takes input vector representations of the words. Also known as embeddings.

In the next decade, neural networks conquered the domain with several breakthroughs.

2014 - the first introduction of sequence-to-sequence models (encoder-decoder). Machine translation counted on that approach.

An encoder network processes a sentence symbol by symbol and compresses it to vector representation. Then, a decoder network predicts the output step by step based on the encoder state.

2018 - first pre-trained language models. It’s a system that is pre-trained on large corpora in a self-supervised fashion. Those pretrained embeddings can be used as featutes in target models or a language model can be fine-tuned.

The rest we know firsthand - large language models now are the most popular NLP representations that help us in many ways.

#ai #nlp #datascience

What is data science?

Even if I've been using data to solve problems for more than a decade, I entered data science just four years ago.

I struggled at first to understand how data science got so popular. Many people talk about it but just a few practice it.

Here's my brief intro to the field and my perspective on its future.

Data science is all about making sense of complex data. It's a strange but effective combination of statistics, computer science, AI, and domain expertise.

It all started with statistics but eventually, computers took over the world and gave statisticians superpowers.

The origins of data science can be traced back to the 1960s-1970s when organizations first started to collect and store large amounts of data.

Later on, personal computing came in during the 1980s, and modern data warehousing architectures became popular.

Back in the 1990s, the Internet changed everything. Communication between all those computers became possible. Millions of websites started generating data.

The rise of social networks in the 2000s boosted the generation of data with billions of users using social networks.

That was the period of distributing computing power and big data. The world entered a new age.

The term “data scientist” was first coined in 2008. It refers to a professional who combines expertise in statistics, programming, and domain knowledge to analyze and interpret complex data sets.

The 2010s brought structure to the field. Data science became a hot topic and suddenly being a data scientist became the sexiest job ever.

The decade that followed made data scientists superstars, and big corporations started investing massively in data science capabilities.

Today, data science is a mature field that evolved a lot. Recent Generative AI breakthroughs changed the focus to ML/AI engineering, but the workhorse of all data-driven improvements is still data science.

Data science is not reserved for big tech giants anymore. With cloud computing an abundance of data and open-source software, anyone can turn insights into actions.
I believe that data science pivots from RnD to more practical use cases. Closing the gap between business and technology is what counts.

A new term called business scientist is emerging - a person who uses data and technology to solve business problems.

Yes, playing with chatbots backed by Large Language Models is fun, but there are so many more important problems that every business faces.

From understanding customers, through forecasting, predicting churn, route optimization, and many more.

Data science is now business science - solving problems by making sense of complex data.

#datascience #businessscience

Deep learning - why do we count on it so much? 

The abundance of computer power enabled what we see today - the deep learning total dominance in AI applications.

Deep learning is a multi-layer artificial neural network.

From self-driving cars to robotics, and text generation, everything is powered by this technology. But why? 

Artificial Neural networks were inspired by how our brain works, and they evolved with time.

Even if they count on huge computer power and immense amounts of data, they prove to give astonishing results. 

The most common neural networks that power contemporary applications are CNNs, RNNs, Autoencoders, Transformers, and other types. 

CNN (Convolutional Neural Network).

The power horse of object detection (image and video).

The architecture of a CNN is designed to mimic the connectivity pattern of neurons in the human brain. It consists of multiple layers that process and transform the input to produce a desired output. 

RNN (Recurrent Neural Network).

The text/speech wizard. That network revolutionized the Natural Language Processing domain.

It's a network that can process sequential or temporal data. RNNs can use their internal state/memory to remember past inputs and generate predictive results.

Main use cases: Translation, text generation. speech recognition and more. For example, given a sequence of words, RNNs predict the likelihood of the next word. 

Autoencoders and Transfomers 

Autoencoders - consist of two main parts: an encoder that compresses the input data and a decoder that reconstructs the input data from the compressed form. Use cases: remove noise from images, compress images, anomaly detection, feature learning, dimensionality reduction. 

Transformers - a different type of architecture, particularly known for their self-attention mechanism, which allows them to handle sequential data effectively, making them suitable for tasks like machine translation and text summarization. 

Use cases: text generation, sentiment analysis, translation, classification, and more. While the original transformer model was an autoencoder with both an encoder and a decoder, modern implementations like OpenAI’s GPT series use only a decoder.

Transformers can be seen as a technique to improve autoencoders, but they are not the same.

To sum up

Artificial neural networks (deep learning) power our modern tech world.

Different types evolved with time, and we count today mostly on CNNs for video/image, RNNs for text, and Transformers for modern apps like ChatGPT.

#ai #datascience #deeplearning

Taplio

Plamen Rabadzhiyski's Linkedin Analytics

Plamen Rabadzhiyski

What is Plamen talking about?

Who is engaging with Plamen

Plamen Rabadzhiyski's Best Posts (last 30 days)

Want to drive more opportunities from LinkedIn?

Taplio

.css-1jgreyo{background-image:linear-gradient(284deg, rgba(39,43,72,1) 0%, rgba(33,47,183,1) 100%);color:transparent;-webkit-background-clip:text;background-clip:text;}Plamen Rabadzhiyski's Linkedin Analytics

Plamen Rabadzhiyski

What is Plamen talking about?

Who is engaging with Plamen

Plamen Rabadzhiyski's Best Posts .css-qqfgvy{font-size:var(--chakra-fontSizes-md);}(last 30 days)

Want to drive more opportunities from LinkedIn?

Plamen Rabadzhiyski's Linkedin Analytics

Plamen Rabadzhiyski's Best Posts (last 30 days)