Program Operation Performance
- KAIST Startup-ting X Spacewalk Review
- 창업원
- 2022-10-30 18:06:48
- 358
https://youtu.be/Lpf9x30miAs
Due to COVID-19, 2020 KAIST Startup-ting was held online through Youtube live streaming. The first company to join us for Startup-ting was Spacewalk, a Proptech startup. Spacewalk provides an AI-based ‘Landbook’ service for land utilization, and has signed MOUs with public institutions such as NH bank.
AI & Proptech – CTO Daniel Lee
The first lecture of Spacewalk’s Startup-ting was about AI and Proptech by CTO Daniel Lee. Proptech is a coinage for the words ‘Property’ and ‘Technology’, referring to the field incorporating technology to real estate (land, buildings, etc.). The Proptech market includes concepts of smart real estate, real estate fintech, contech, and collaborative economy. The field is anticipated to solve problems between the market and customers through its technology and platform. Whereas sharing economies like Airbnb are Proptech v 2.0, companies that use big data or technology to help existing markets are v 3.0, to which Spacewalk belongs. Currently in Korea, asset types (residential, commercial, etc.) and service types (brokerage, information provision, construction, management, sharing, finance, etc.) combine with technology to form various fields in Proptech.
Asset managers and experts usually deal with large-scale real estate, but Spacewalk helps people build optimized buildings through technology and automation. Complex estate laws make profit estimation difficult and cause large deviation between buildings in the same area. Since estate is an area difficult to deal without experts, Spacewalk helps customers by predicting profits before and after real estate development. Spacewalk provides information (value evaluation) to landowners and investors, provides liquidity (purchase consultation and asset management model), and forms interests with real estate agents, consultants, public institutions, and architects.
Spacewalk uses deep learning to solve nonlinear problems, and has a Tech Pipeline that uses public data to engineer data, construct algorithms, and go through machine learning to derive results. Spacewalk owns and develops the world’s finest technology through deep reinforcement. Using artificial intelligence, the company generates the building’s exterior, plans parking space, allocates land use, and predicts value using an automated value evaluation model. Landbook, a software service provided by Spacewalk, reviewed about 500,000 m2 of land, estimated to be about 1.5 trillion KRW. There are more than 18,000 monthly active users and at least 1,500 daily active users of Landbook. Experts and masters of computer science, physics, mechanical engineering, and architecture work at Spacewalk to solve real estate problems and pursue expertise and diversity.
The Real Work Story of Proptech Developers and Engineers I – Full-stack Development, the misunderstandings and truth – Web Engineering Manager Wonmin Jeon
The second lecture is about the real work stories of Proptech developers and engineers by 3 developers. Wonmin Jeon, the Web Engeering manger, gave the first lecture about developing a Full-stack web.
Full-stack web development is made up of Front-end (Screen implementation & management), Back-end (Server implementation & management using framework made of languages like php, python, ruby, c#, etc.), and Devops (on-primise or cloud system construction & management). Web developers are required to have skills such as Git, Basic Terminal Usage, Data structures & Algorithms, GitHub, Licenses, Semantic Versioning, SSH, HTTP/GTTPS and APIs, Design Patterns, and Character Encodings. Front-end is web-based so it requires understanding of the internet, and Back-end requires understanding of communication and various skills, and knowing different programming languages.
Front-end is relatively simple because it communicates by receiving and manipulating data, but Back-end requires a deep understanding of algorithms to store, edit, delete and manage data. Devops is a stage required to understand which systems products are based on, and to deal with traffic. It is not easy to excel in all stages, so it is important to select a development method depending on the size of the startup. If the startup is at the beginning stage, it should choose a stack to use, and then develop a product and go through trial and error. If the company is the size of a unicorn company, it already has experts for each stack, and the stack depends on the organization’s situation.
To become a full stack developer, one must become a generalist, not a specialist. It is recommended to know Front-end well, and learn Back-end skills depending on the company’s characteristics. Devops requires basic knowledge of networks, linux, windows, OS, etc., and Cloud knowledge is mandatory. However, recognizing and solving problems as well as communicating with people are as important as having the skills.
The Real Work Story of Proptech Developers and Engineers II – Data Science at Spacewalk – Data Scientist Juhwan Hong
With the advent of Proptech 3.0, data science can be divided into three main fields: Smart Buildings, Collaborative economy, and Real estate Fintech. Firstly, in ‘Smart Buildings’, real estate management automates building control using sensor technology and IOT-based big data, and optimizes energy consumption. In ‘Collaborative economy’, such as Airbnb and Wework, it is important to process user data that stacks up every second. Finally, the ‘Real estate Fin Tech: Prop + Fin Tech’ field automates all real estate evaluation, including land and buildings, such as brokerage and leasing services, and investment and financing crowdfunding services. Spacewalk also works in each field for land utilization.
Real estate appraisal can be divided into single-property appraisal and mass appraisal. Mass appraisal is the process of valuing a group of properties using statistical models, and the demand for such is increasing. Various Automated Valuation Models (AVM) can be used in areas such as evaluation of mortgage (housing mortgage loan and estimating mortgage value for real estate-related financial products), determination of the amount of the tax base (including property tax, general income tax, inheritance tax, gift tax, etc.), economical aspects (identifying and predicting changes in supply in the land and housing markets, and establishing policies), financial management aspects (asset portfolio managing including real estate), and real estate development evaluation (business value assessment through value computation before and after development).
Real estate development evaluation is the service currently provided by Spacewalk. Spacewalk is planning a business model that will pre-emptively capture places with high business value through AVM by combining future development assessment with economical and financial management aspects. Since there is a limit to the traditional mass appraisal methods, Spacewalk is developing a non-parametric AVM using data science. Without standardizing real estate value into a specific function, Spacewalk is exploring various functions and finding a value function based on data that minimizes the researcher’s subjectivity. A non-parametric AVM can utilize various non-parametric model-based machine learning algorithms. Spacewalk goes through the following data science process to develop an AVM – 1. Frame the problem, 2. Collect raw data, 3. Process the data, 4. Explore the data, 5. Perform in-depth analysis, and 6. Communicate results. The roadmap to developing the AVM can be divided into three dimensions – space (current coverage range (Seoul, Gyeong-gi, Incheon, Pusan)-> expansion across the country->worldwide), time (future value, long-term plans), and usage (small residential real estate -> commercial, agricultural land, etc.). Spacewalk’s AVM will go beyond a simple valuation model to be used in optimizing real estate investment portfolios and use a deep reinforcement training module to optimize land utilization.
The Real Work Story of Proptech Developers and Engineers III – Constructing a flexible and expandable container-based big data pipeline – Data Engineering Manager Junpyo Lee
Followed up, Data Engineering Manager Junpyo Lee gave a lecture about constructing a flexible and expandable container-based big data pipeline.
Data Engineering is systematizing (automating) the process of moving source data to the target according to the business logic. Data pipeline, also known as ETL, refers to the set of processes of Extracting data from the internal DB, external server, or cloud through APIs or crawling, Transforming it, and Loading it into some database. To systematize a pipeline, a pipeline management tool is necessary, and to operate a pipeline, containers are used. Containers offer a logical packaging mechanism in which applications can be abstracted from the environment in which they actually run. Containers virtualize at the operating system level, so they are lightweight, start much faster, and use a fraction of the memory to boot. In addition, containers are consistent no matter where the application is deployed. Container management tools virtualize single or multiple systems, provide a container compatible environment, monitor containers, as well as reboot, expand, and copy container system resources. In Spacewalk, we extract source data, and load it to field data (real estate, finance, etc.), and service it using Landbook.
Spacewalk creates data pipelines for flexibility. In data pipelines, there are a variety of development entities (DS, DE) that develop Tasks. Each task has different libraries and associated dependencies required to run. So, container-based data pipelines were used to provide a stable environment and process to distribute and test tasks. In addition, due to the increase in service products and source data, the workflow and number of tasks are increasing, so it was necessary to create an expandable computer resource.
A Q&A session followed the lecture with real-time questions and questions received in advance. The speaker corresponding to the topic of the question resolved the curiosity of the participants. Spacewalk gave explanations and advice on topics from how startup culture can settle in a manufacturing business, whether data on value assessment of real estate locations exists, how Spacewalk’s technical aspects apply to business, to specific examples of big data pipelines.
We hope you learned about AI, Proptech, and startups from Spacewalk! It is unfortunate that we cannot meet face-to-face, but thank you for joining and communicating with Spacewalk online. Please look forward to June’s Startup-ting ????
- 다음
- August 2019 KAIST Startup Awards Review
- 2022-10-30