Blog by Jae W. Lee

Online reviews submitted by customers and end users have a significant influence on online businesses. According to a study, online reviews can affect a customer’s purchasing decision and can also impact a business’s overall online reputation.

Article: Closed-form evaluations and open-ended comment options: How do they affect customer online review behavior and reflect satisfaction with hotels?

The referred article discusses how different forms of customer online reviews (closed-form evaluations vs open-ended comments) affect customer behavior and reflect satisfaction with hotels. Closed-form evaluations are structured reviews that ask customers to rate certain aspects of their hotel experience on a scale (e.g. cleanliness, comfort, etc.). Open-ended comments, on the other hand, allow customers to write a free-form review of their experience.

The article argues that closed-form evaluations may be more useful for hotels to identify specific areas for improvement, while open-ended comments may provide a more detailed and nuanced view of the customer’s experience. Additionally, the article suggests that customers may be more likely to leave a review if they are given the option to leave open-ended comments, as it allows them to express their feelings more fully.

In summary, the article suggests that closed-form evaluations and open-ended comments have different effects on customer behavior and satisfaction, and may provide different insights for hotels looking to improve their guests’ experiences.

Designing a data lake on Amazon Web Services (AWS) and Google Cloud Platform (GCP) both have their own unique set of features and services. 

First, let’s discuss designing a data lake on AWS. AWS provides a variety of services that can be used to build a data lake, such as Amazon S3 for storage, Amazon Glue for data cataloging and ETL, and Amazon Athena for SQL querying.

To design a data lake on AWS, you can start by creating an S3 bucket to store your raw data. Next, use Glue to create a data catalog, which allows you to easily discover, understand, and connect to your data. Glue can also be used to perform ETL on your data, so you can prepare it for analysis. Finally, use Athena to query your data using SQL.

Now, let’s talk about designing a data lake on GCP. GCP also provides a variety of services that can be used to build a data lake, such as Google Cloud Storage for storage, Google Cloud Dataflow for data processing and ETL, and BigQuery for SQL querying.

To design a data lake on GCP, you can start by creating a Cloud Storage bucket to store your raw data. Next, use Dataflow to perform ETL on your data, so you can prepare it for analysis. Finally, use BigQuery to query your data using SQL.

Overall, both AWS and GCP provide similar services for building a data lake. However, the specific services and tools used may vary. For example, AWS provides Glue for data cataloging and ETL, while GCP provides Dataflow for data processing and ETL. Additionally, AWS uses Athena for SQL querying, while GCP uses BigQuery. Both services are powerful and flexible, so the choice of which one to use will depend on your specific use case and requirements.

In case you need a data lake without any other third-party dependencies, AWS provides all the services needed to design a data lake. Similarly, GCP provides all the services needed to design a data lake.

In summary, designing a data lake on AWS and GCP both have their own unique set of features and services. Both are powerful and flexible, so the choice of which one to use will depend on your specific use case and requirements. However, AWS provides Glue for data cataloging and ETL, while GCP provides Dataflow for data processing and ETL. Additionally, AWS uses Athena for SQL querying, while GCP uses BigQuery.

A Customer Data Platform (CDP) is a platform that helps businesses manage and extract insights from customer data. It is a central repository of customer data that can be accessed and utilized by various departments within a company, including marketing, sales, and customer service. CDP platforms consist of tools such as a customer database, marketing automation, and management tools for multichannel campaigns and real-time customer interactions.

One of the primary benefits of a CDP is that it allows businesses to have a comprehensive view of their customers. By centralizing customer data from various sources, such as website interactions, social media interactions, and customer service inquiries, businesses can gain a more holistic understanding of their customers. This can be especially useful for businesses that rely on a multi-channel marketing approach, as it allows them to personalize their marketing efforts based on a customer’s specific interactions with the company.

In addition to providing a more comprehensive view of customers, CDP platforms also make it easier for businesses to manage and analyze customer data. With the right tools, businesses can segment their customer base, identify key trends and patterns, and create targeted marketing campaigns. This can help businesses improve the effectiveness of their marketing efforts and increase customer loyalty.

While building a CDP from scratch can be a daunting task, there are a number of companies that offer CDP platforms that businesses can utilize. One such company is Salesforce, which offers a CDP platform that integrates with its other marketing and sales tools. This can be a useful option for businesses that are already using Salesforce and want to expand their data capabilities through a 3rd party tool.

However, there are also pros and cons to expanding data capabilities through 3rd party tools and platforms. One of the main pros is that it can be more cost-effective than building a CDP from scratch. It can also be easier to implement and maintain, as the platform is managed by the vendor. On the other hand, there may be limitations to the customization and integration options available with a 3rd party platform, and businesses may have to pay a subscription fee to use the platform.

Regardless of whether a business decides to build their own CDP or use a 3rd party platform, it is important to define the Single Source of Truth amongst the system architecture and business platforms. This means ensuring that all data is consistent and accurate across all integrated 3rd party solutions. This can be challenging, as different systems may have different data structures and definitions, but it is essential for ensuring that customer data is reliable and can be trusted for decision-making purposes.

In conclusion, a Customer Data Platform (CDP) is a valuable tool for businesses that want to manage and extract insights from customer data. It can be integrated with a CRM system to gain a more comprehensive view of customers, or with an email marketing automation tool to send targeted, personalized emails. While there are pros and cons to expanding data capabilities through 3rd party tools and platforms, defining the Single Source of Truth is key to ensuring that customer data is accurate and reliable. Potential business cases where a CDP can empower online-centric businesses include Software as a Service (SaaS) products, online learning platforms, ride share and food delivery businesses, and even customer review platforms, all of which can benefit from higher user retention and a personalized user experience.

Generative AI is a type of artificial intelligence that is focused on generating new content or data. This can be in the form of text, images, audio, or any other type of media. Generative AI is achieved through the use of machine learning algorithms and various modeling techniques, including predictive models.

The creation of a Generative AI platform begins with the collection and preparation of data. This involves gathering a large dataset of examples that the Generative AI model can use to learn from. This dataset should be diverse and representative of the type of content that the Generative AI platform is intended to generate. For example, if the platform is meant to generate music, the dataset should include a wide variety of music from different genres and styles.

Once the dataset has been collected and prepared, the next step is to design the machine learning algorithms and models that will be used to train the generative AI platform. There are several different types of machine learning algorithms and models that can be used for this purpose, including supervised learning algorithms, unsupervised learning algorithms, and reinforcement learning algorithms.

Supervised learning algorithms are trained using labeled data, where the correct output is provided for each input. These algorithms learn to predict the output based on the input data, and are commonly used for tasks such as classification and regression. Unsupervised learning algorithms, on the other hand, are trained using unlabeled data, and are used to discover patterns and relationships in the data. These algorithms are commonly used for tasks such as clustering and dimensionality reduction. Reinforcement learning algorithms are trained using a reward system, where the algorithm receives a reward for performing a desired action. These algorithms are commonly used for tasks such as controlling robots or playing games.

Once the machine learning algorithms and models have been designed, the next step is to train the Generative AI platform. This involves feeding the dataset into the machine learning algorithms and models, and adjusting the parameters of the algorithms and models to optimize their performance. This process is known as “training,” and it is typically done using a combination of manual tuning and automated optimization techniques.

After the Generative AI platform has been trained, it can be used to generate new content or data. This is done by providing the platform with a prompt or seed, which serves as the starting point for the generation process. The Generative AI platform then uses its trained machine learning algorithms and models to generate a response based on the prompt. The response is then fed back into the platform, which uses it to generate the next response, and so on, until the desired amount of content or data has been generated.

The role of machine learning algorithms and modeling in Generative AI is crucial, as they are responsible for learning the patterns and relationships in the data and using this knowledge to generate new content or data. Predictive models, in particular, play a significant role in Generative AI, as they are used to predict the likelihood of certain outcomes based on the input data. These models can be used to generate responses that are more likely to be relevant or appropriate based on the prompt or seed.

Artificial intelligence is achieved in Generative AI through the use of machine learning algorithms and modeling techniques, which allow the platform to learn from data and generate new content or data based on this learning. The resulting content or data is typically of high quality and can be used for a wide range of applications, including creative writing, music generation, image generation, and more.

While Generative AI has the potential to bring many benefits, it is important to consider the role that data plays in the success of these systems. Ergo, the quality and uniqueness of the data used to train Generative AI systems is critical to their performance. If the data is poor quality or not representative of the real world, the AI system may not be able to generate accurate or useful content.

Overall, Generative AI is a powerful tool that has the potential to revolutionize the way we create and consume media, and will likely continue to play an increasingly important role in the world of artificial intelligence.

The search-inference-framework is a cognitive model that explains how humans think and make decisions. It is based on the idea that we resolve doubts and make decisions by thinking about possibilities and searching for evidence to support our judgement. This process is driven by our beliefs and goals, and it is influenced by various factors such as our prior knowledge, our emotional state, and the context in which we are making the decision.

According to the article "The search-inference-framework of thinking" by Bernhard Wenzel, the search-inference-framework consists of three main components: search, inference, and evaluation. The search component refers to the process of actively seeking out and gathering information that is relevant to the decision at hand. The inference component refers to the process of using this information to form a belief or opinion about the decision. Finally, the evaluation component refers to the process of weighing the pros and cons of different options and deciding on a course of action.

One key aspect of the search-inference-framework is the role of a large data set, which can be thought of as a dictionary or encyclopedia of information that we can use to inform our decisions. This data set is not only useful for finding solutions to specific problems, but it can also be used to train machine learning models, which can then be used to improve decision-making processes. For example, a machine learning model trained on a large data set of customer data might be able to predict which marketing strategies are most likely to be successful for a particular business.

Artificial intelligence has created new opportunities to improve the search and discovery process for finding solutions to problems. One way that AI has done this is by automating many of the tasks involved in search and inference, such as data gathering and analysis. This has allowed humans to focus on more complex tasks, such as evaluating different options and making decisions. In addition, AI has also made it possible to process and analyze large data sets much more quickly and accurately than humans could, which has led to more efficient and effective decision-making.

Overall, the search-inference-framework is a powerful model for understanding how humans think and make decisions. It highlights the importance of gathering and analyzing relevant information, as well as the role of prior knowledge, beliefs, and goals in shaping our decisions. Artificial intelligence has created new opportunities to improve this process by automating many of the tasks involved in search and inference, and by allowing us to process and analyze large data sets more efficiently.

Reinforcement learning is a type of machine learning that involves training algorithms to make decisions in a dynamic environment. It is based on the idea of an agent learning to interact with its environment in order to maximize a reward. This approach has been successfully applied to a variety of domains, including robotics, natural language processing, and video game playing.

One way that reinforcement learning could be used in an e-commerce platform is to optimize the recommendations that are made to customers. Traditional recommendation algorithms rely on collaborative filtering, which uses the past behavior of similar users to make recommendations. However, this approach does not take into account the specific context of the individual user or the real-time availability of products.

Reinforcement learning, on the other hand, can learn to make recommendations based on the current context of the user and the current state of the inventory. For example, if a customer is browsing a particular category of products, the reinforcement learning algorithm could learn to recommend related products that are currently in stock and have a high probability of being purchased. This could increase the likelihood of making a sale and improve the overall customer experience.

Another way that reinforcement learning could be applied to an e-commerce platform is in the optimization of pricing. Setting the right price for a product can be challenging, as it needs to be high enough to cover the costs of production and distribution, but low enough to be competitive and attract buyers. A reinforcement learning algorithm could be trained to adjust prices based on the demand for a particular product and the competition from other sellers.

For example, if a product is in high demand and there are few similar products available, the algorithm could learn to increase the price in order to maximize profits. On the other hand, if the demand for a product is low or there are many similar products available, the algorithm could learn to decrease the price in order to increase sales. By continuously adjusting prices in this way, the algorithm could help the e-commerce platform to optimize its profits.

In summary, reinforcement learning has the potential to be applied in a number of ways to an e-commerce platform. It can be used to optimize recommendations to customers and to optimize pricing in order to maximize profits. By continuously learning from its interactions with the environment, a reinforcement learning algorithm can adapt and improve over time, leading to better results for the platform.

Reinforcement learning and modeling are powerful techniques that enable businesses to become data driven by allowing them to optimize their decision-making processes based on data-driven insights. These techniques are based on the idea that actions taken by a system or machine can be reinforced or punished based on their outcomes, leading to the development of optimized behaviors over time. Through the use of reinforcement learning and modeling, businesses can improve their performance, increase efficiency, and deliver a more personalized user experience.

One of the most well-known examples of a business that has successfully implemented reinforcement learning is Netflix. The company’s recommendation system is powered by a combination of collaborative filtering and reinforcement learning, which enables it to deliver personalized recommendations to each of its users based on their past viewing history and other data points. According to a report from Wired, this system has been a key driver behind Netflix’s success, leading to an increase in stream time of up to 80%.

Another example of a business that has used reinforcement learning to great success is Tesla. The company’s autonomous driving technology is powered by a combination of machine learning and reinforcement learning, which allows its vehicles to learn from their experiences on the road and make data-driven decisions about how to navigate traffic and other hazards. This has enabled Tesla to develop one of the most advanced autonomous driving systems in the world, and has helped the company to become a leader in the field of electric and autonomous vehicles.

Another company that has implemented reinforcement learning in its operations is Spotify. The music streaming giant uses a recommendation algorithm that is powered by reinforcement learning, which allows it to deliver personalized recommendations to each of its users based on their listening history and other data points. This has been a key driver behind Spotify’s success, helping the company to differentiate itself from competitors and retain a large and loyal user base.

There are several key benefits that businesses can derive from the use of reinforcement learning and modeling. First and foremost, these techniques allow businesses to optimize their decision-making processes based on data-driven insights, rather than relying on gut instincts or preconceived notions. This can lead to improved performance and increased efficiency, as businesses are able to make more informed decisions that are based on real-world data rather than assumptions.

In addition, reinforcement learning and modeling allow businesses to deliver a more personalized user experience. By analyzing data about individual users and their behaviors, businesses can tailor their products or services to better meet the needs and preferences of each user. This can lead to increased customer satisfaction and loyalty, as well as higher levels of engagement and retention.

Overall, the use of reinforcement learning and modeling is a powerful tool for businesses that want to become data driven. By optimizing their decision-making processes and delivering a more personalized user experience, businesses can unlock the full potential of their data and stay ahead of the competition. The case studies of Netflix’s recommendation system, Tesla’s autonomous driving technology, and Spotify’s recommendation algorithm demonstrate the many benefits that businesses can derive from these techniques, and provide a roadmap for other companies looking to implement them in their own operations.

For myself, I have found it helpful to make a conscious effort to stay hands-on and maintain my technical skills. One way I have done this is by regularly sharing short videos introducing sample python code with the public through a YouTube playlist.

Not only does this help me stay up-to-date with new developments in the field, but it also allows me to stay creative and challenge myself to think about problems in new ways.

In addition to the personal benefits of keeping my technical skills sharp, there are also practical advantages to being able to “ride the horse” as a leader. Having a strong foundation in the technical aspects of my work allows me to better understand the challenges and opportunities facing my team, and to effectively communicate with and support them. It also allows me to be a more credible and effective leader, as I am able to directly contribute to the work being done rather than just overseeing it from a distance.

In conclusion, it is important for leaders to remember the importance of maintaining hands-on skills and not becoming too removed from the practical aspects of their work. As the saying goes, “don’t be a cavalry captain who can’t ride a horse.” By staying grounded in the technical skills that got us to where we are today, we can better understand and support our teams, and continue to be effective and credible leaders in our field.

머리에는 다저스 모자 눌려있고

가슴에는 아이팟을 걸고 있으며

별다방이나 던킨에서 모여 앉아

외국생활 경험을 이야기하거나

외국말을 얼마쯤 지껄이는 자가

어찌 글로벌시대의 인재라 할 수 있겠는가

이는 글로벌시대의 실패작도 아니고

글로벌시대의 쓰레기도 아니다

Globalization이란 헛바람에 날려서 마음속에

주견도 없는 한낱 글로벌시대의 병신이다.

- 2007/05/18 17:37 <이재왕의 디지털 錄> 중에서

For startups, interviewing software engineers is as hard for hiring managers as it is for jobseekers. That is because it is a two-way street. Startups are looking for a very particular type of talent. On the other side, there are specific elements that lead talent to want to work for a startup in the first place. Engineers interviewing with startups are often looking for a particular type of organization, role, and workplace culture. However, a lack of clear information often makes this process more complicated than necessary. The more startups understand what drives candidates to consider this type of employment, the better chance they will have to find the right fit.

My view on building the right team and a thriving company have changed over time. That has been due to multiple failures and a few lucky wins. I recently had a chance to reflect on all this with other startup leaders.

"Build the best team possible."

I recently received an invitation to join a Clubhouse event focused on the topic of startups and hiring. Over 250 people listened in with speakers including early days ex-Uber launchers and a friend of mine who created the very first AT&T developer relations program. When the floor was opened for discussion, we all got to hear several different views on hiring from the perspective of candidates.

There was much reflection on the question of what a candidate looks for when choosing a startup to interview with and eventually work for. Several factors were mentioned that potential recruits look at. They included the TC (total compensation), company culture, and the background of the founders. Following that, some of the event's speakers and audience members who were hiring managers shared from the other perspective. What became clear is that finding better ways to communicate the right information between candidates and hiring managers may better bridge the gap in the interview and hiring process.

"How far does a startup go to hire top talent?"

Obviously, not all startups are funded with venture capital. So not every hiring manager has access to an unlimited expense account or the backing of a People and Talent team. Not every recruitment effort needs that level of finance and effort. Still, are startups doing the right things to find the right fit?

Andrew Kim shared a fascinating statistic with the audience. When searching for new talent, it takes approximately 3-5 seconds for someone from the HR team to screen the first batch of resumes. That is a cursory look compared to the time that some candidates put into perfecting them. Still, I wouldn't be surprised if fellow software engineers spend about the same amount of time skimming through each job opening on LinkedIn. Touché.

What does all this mean?

The two-way street of hiring comes down to communications. Perhaps, for hiring managers, taking the time to draft a unique and personalized job description would go a long way for a startup. Maybe candidates would take more than a few seconds to review it if they believed it wasn't another carbon-copied template. As for the job seeker, making their resume more useful to the hiring manager by highlighting specific problems and challenges one has solved in her/his career could help. Apparently, video resumes are becoming more popular and can add a new dimension to the review process.

Alex Donn reminded the audience of something that goes easily forgotten for candidates: know what you value. This is critical before beginning the search process, and it is vital throughout the hiring process. Your values are quite different from your market value. Ironically, I believe the latter is more spoken about and discussed. Understanding your own values prepares you to make a better decision about the jobs you seek to pursue. Also, candidates with clear values make the role of the hiring manager a lot clearer, especially for startups.

Great opportunities find top talent - and vice versa. When these two meet, I believe it is due to the alignment of vision, value, and priorities. Startups and engineers need to understand their own unique values. Startups are relatively risky, but candidates know that. They also know that startups create a unique opportunity for growth. Great startups need the talent to get things done and also to make critical decisions. As a result, top talent is given a chance to make complex and mission-critical decisions and grow at a rate not always possible at an established organization.

Yes, startups have risks, but they can also provide comparable rewards, and candidates know that risks also apply to joining a corporation if you are not clear about what you are seeking. Still, if candidates are simply focused on finding a job with better pay and more flexible office hours, they may not find their home with a startup.

Find your rocket ship.