Akifumi Kobashi After graduating from the University of Tokyo, he joined McKinsey & Company. As a Senior Manager at the firm, he supported major manufacturers in various industries such as equipment, machinery, heavy industry, and transportation in global purchasing and procurement reforms. In 2017, he founded CADDi Inc. He majored in Electrical Engineering at Stanford University and Graduate School, and worked at a U.S. aerospace and aircraft manufacturing company while studying. After engaging in semiconductor security enhancement at Qualcomm, he led battery sustainability improvements and the development of the sensor components for AirPods at Apple U.S. headquarters. In November 2017, he co-founded CADDi Inc. with Kato Employees (global) Raising capital Global offices 21.7 Billion JPY 4countries 2024年8⽉1⽇時点 Japan, USA, Thailand, Vietnam Overview Founders 600pax As of Aug 2024
We aim to create a society where everyone involved in manufacturing can fully realize their inherent potential. To achieve this, we will establish "new mechanisms" that transform the conventional wisdom of the industry. In the current manufacturing sector, a vast amount of potential remains untapped. Many are overwhelmed by estimation and management tasks, lacking sales capabilities, information, and networks. These constraints prevent them from showcasing their true development and technical abilities. By easing these restrictions, we can unlock the potential of each company, generating significant strength across the entire industry and enriching it. From small local factories to large, historically significant manufacturers, and newly established ventures—every manufacturing company can leverage its strengths to shine and create new value. We will continue to strive to carve out such a future.
The company name “CADDi” combines the meaning of “ordering directly from CAD” and “the desire to become an infrastructure that supports players in the manufacturing industry,” likened to a caddy in golf.
phase" 7 2017 2018 2019 2020 2021 2022 2023 2024 Kansai Branch Office opened Head Office expansion and relocation Founded on the 150th anniversary of the Meiji Restoration CADDi Drawer started Established the corporation in Thailand Established the corporation in Vietnam Head Office expansion and relocation We've just begun the second chapter as a "Manufacturing AI Data Platform " Series B raised 8.03 billion JPY Established the U.S. corporation Series C raised 11.8 billion JPY Business Integration
that support our daily lives The manufacturing industry is involved in all the things we need in our daily lives, including medicine, food, transportation, and communications. All kinds of things are produced by the hands of many people, including industrial machinery, machine tools, and processing companies. Smartphones consist of more than 1,000 components, including semiconductors, batteries, and displays Industrial Features
work and data are not utilized, and the same things are done over and over again from scratch every time. "Struggle and learn" Because I had to struggle too Reinvention of the wheel plaguing the Manufacturing Industry Design Procurement Sales Creating drawings similar to past ones Starting from scratch with multiple quotations and negotiations Handling similar quality defects Estimate work from scratch Despite being the largest industry, innovation is not happening Dependence on veterans, faxes and phone calls remain in the field Industry is nearing its limits Operations are not becoming more efficient even as the population declines, and the industry as a whole is nearing its limits Same way of doing things as decades ago A string of supplier bankruptcies Inefficiency on the site The number of small and medium-sized manufacturers(suppliers), has decreased by about half over the past 20 years Manufacturing & Quality Assurance
wheel due to data fragmentation 11 調 達 製 造 販 売 部署や担当者、会社を跨いで、 各⾃バラバラにデータが管理さ れ分断されている。 Data Silos Fixed translation and closed relationships are refined within the same environment. Individual Optimization Due to a long value chain and numerous stakeholders, individual optimization progresses. Data is managed separately and fragmented across departments, responsible parties, and companies. procurement Manufacturing 11 Previous work and data are not utilized, and the same things are done over and over again from scratch every time. "Struggle and learn" Because I had to struggle too Reinvention of the wheel plaguing the Manufacturing Industry Design Procurement Manufacturing & Quality Assurance Sales Creating drawings similar to past ones Starting from scratch with multiple quotations and negotiations Handling similar quality defects Estimate work from scratch Industrial Features 設 計 design Sales
aiming for and our approach Capitalize on all operations and data, and automate and optimize them with AI and technology. A world where more innovation is generated throughout the manufacturing industry. Data is not capitalized Data is capitalized. Leverage" the accumulation of the past Accelerate growth through standardization Connect all siloed data and organize them into a usable state Repetition of the same task Reinventing the wheel
in the manufacturing industry. We do not just store them, but utilize them as “assets” that will create value in the future. Drawing out what has been stored and creating value. CADDi Drawer” is a platform for this purpose. Drawing Drawer CADDi Drawer Application ①
Analysis 15 Surface Roughness Indicates the smoothness of the finished product. Higher smoothness requirements lead to increased costs. Chamfering Refers to the process of beveling edges, which can be done at angles like 45° or using rounded methods.The choice of machinery impacts pricing and sourcing Geometric Tolerances Precision and allowable deviations of part shapes; smaller deviations make finding capable manufacturers harder. Symbols indicate specifications, and numbers show tolerance ranges. Hole (Φ) It may not be readable by third-party OCR Title Block There is a variety of information tied to the drawing. We want to get it in key/value pairs. Blanks Fairly sparse Handwritten Txt the inconsistency in style and legibility. Rotated Characters Characters rotated by 90° or presented vertically may not be read correctly. Dimensions Dimensions may be listed in both millimeters and inches It reflects similarity not just in appearance but also in the actual shape of the item. Similarity Finished Product Image Application ①
of "drawing information" into digital data and make it freely accessible. CADDi Drawer transforms drawing data into assets Drawings etc.. quality Design Estimation Orders Received Orders Placed Store Drawings Search Extract Information Associate Information Take Action Analyze Feature 1 : Automatic Drawing Analysis Automatically analyze text information and product shapes within drawings Feature 2 : Keyword Search Search drawings using any text information, including handwritten notes within the drawing, as keywords Feature3: Shape search Utilize the analyzed product shape data to conduct similar shape and image searches Feature 4: Link Necessary Information Link related information, such as order details, defect data, and 3D models, to 2D drawings Patent Obtained Application ① After suppressing the most important drawing data, go on to connect all the peripheral data.
manufacturing industry, where costs account for 80% of total sales, the "quotation process" for determining prices is extremely important. However, until now, there has been a lack of significant innovation in this area, and many problems have been faced, such as the delegation of work to veteran employees and the tightening of workloads due to e-mail and fax-based implementation. “CADDi Quote" will greatly improve the quotation experience and create a cycle in which daily work becomes an asset. AI Quote Cloud
Costs Based on Similarities Utilizing past performance and similarity searches to propose the most suitable supplier candidates. Similarity-based target cost setting Attach price performance data and refer to other information stored in the Drawer. Streamlining Tedious Estimation Tasks Batch creation of estimation requests to multiple suppliers, along with price collection and comparison. あ Transforming estimation data into assets for utilization. Automatically accumulating and analyzing obtained estimation information. How will Quote enable faster and stronger estimations? Application ②
manufacturing business, we have helped solving a variety of problems in manufacturing industry. We will create new applications based on the know-how and technology we have cultivated in this business. We have nearly 100 ideas and plan to develop dozens of applications over the next 3 years. New Products
Started software business Software for Manufacturing Global Top Class Growth Companies that have introduced our services in Japan(examples) Awards Rapid Growth
aiming for 10x growth Compounding Strategy Utilizing the know-how cultivated in the Manufacturing business, dozens of applications will be deployed within three years. Data Gravity Development of functionality/infrastructure that will allow all data that exists in the manufacturing industry to be aggregated within Drawer. Global Strategy Development of killer features/regulatory compliance for high growth in US and other countries. Drawing 3D CAD Specifications Defect Information Quotation Record of proceedings Defective pictures Order amount Data exploration and analysis Advancement of procurement operations Document Summary = Infrastructure for integrating and maintaining information A group of business applications leverage integrated data to create value Chatsearch Application (Utilization) Data Platform (Integration and analysis)
Data Architecture Rapid Growth Global support We are developing products to be widely used in the U.S. and other parts of the world.Each product must be designed on the premise that it is not confined to Japan. Compounding Strategy As the number of applications grows by the dozens, it is necessary to organize the dependencies as a whole while growing each one. • Assumption of 10x annual growth CADDi aims to change the world in the next 10-20 years time horizon. The angle of growth required is unprecedented in Japan. In development, we are forced to make difficult decisions on a daily basis on how much technical debt to tolerate and how much functionality to build in. Existence of the physical world The manufacturing industry is not completed in the digital world. How to seamlessly connect with the real world is an extremely challenging challenge from a data perspective. The number of stakeholders Manufacturing involves hundreds of people just to make one component.The requirements for certification/authorization and UI/UX are extremely complex and sophisticated. • A wide variety of data We handle not only structured data, but also unstructured data such as drawings, product images, and documents.Engineers are tested on how to create customer value using a large volume and variety of data.
4 7 6 Many technical issues remain, and the phase of development is still in progress. Architecture Improvement Boundary setting in a multi-product environment Distributed systems, microservices/modular monolith design DDD implementation, multi-region/multi-tenant design Improve agility of value verification Launch of new products for PMF Abstraction of individual issues and technical incorporation into the platform Scale and agility of the organization/product Improvement & delivery of ML models Significantly shorten the lead time from initial verification to implementation Maintain model performance and improve accuracy in environments where data drift can occur. Defining the ToBe of data management in an environment of increasing data types Definition of product quality Quality assurance of mission-critical systems Quality Definition and Measurement System for Core Values Quantify correctness and value of analytics/AI-based solutions DevEx Improvement Platform engineering Common infrastructure/common libraries for traceability, certification, authorization, etc. Establishment of development guidelines Definition of what UI/UX should be Design in the absence of similar products Designing experiences that straddle multiple products and building design systems Responding to and expressing a wide variety of stakeholders and use cases Ensure scalability of the organization Management of a doubling development organization and training of mid-level leaders Development management of multiple projects Globalization / Cross-cultural and cross-language issues, joint development with Vietnam
drawings and documents to stabilize manufacturing quality. Quotation Recommended Suppliers for Quotation Supplier selection is often based on personal experience and judgment, leading to a subjective process. By selecting suppliers based on past quotation records, processing characteristics from drawings, and manufacturing performance data, we can ensure consistency and transparency, enabling efficient supplier selection and anticipating cost reductions and quality improvements. Future Vision and Research Development Themes 25 Data Related to Manufacturing Preventing risks during the design phase Digital tools for creating and editing 3D models and detailed design drawings CAD (3D) Design information expressed in 2D drawings, including product shape, dimensions, and materials. Drawings Numerical and categorical information managed in standardized formats, such as bills of materials (BOM) and inventory lists. Structured Data Text related to products, including specifications, manuals, and reports. Documents Drawing etc.. Quality Design Quotati on Order Receiv ed Order Place Product Design Preventing risks during the design phase Design reviews are conducted at each stage of design to verify and improve quality, functionality, and cost. By making information about related products and components easily accessible during design review preparation and execution, we can enhance review efficiency and reduce the risk of overlooking critical points. Ensuring manufacturing quality relies on the expertise of veterans and implicit knowledge from past transactions, which is often not clearly shared. By analyzing drawings and specifications to extract processing characteristics, we can identify critical points to watch during manufacturing, reducing risks and stabilizing/improving quality. • Cost information • Supplier information • Customer Specification • Issue History • design philosophy / memo • VA/VE Ideas Are there any issue that have occurred with similar products? Supplier A：¥2,000 Supplier B：¥2,100 Supplier ーF：¥2,050 … ー メ カ ー Bending Punching/ Sheet Thickness minimum geometric intersection ‧‧‧ Risk A Risk B Risk C Risk D Manufacturing Risk Detection Extract Processing Characteristics Drawings/Documents Mach ining Sheet Metal manufacturer
・Apollo Client ・Recoil ・Storybook ・Jest ・Lerna BFF ・TypeScript ( NestJS ) ・Apollo Server Backend ・Rust ( axum ) ・TypeScript ・ Node.js ( Express, Fastify, NestJS ) ・Java, Kotlin *Depending on the product characteristics, other languages may also be used besides those mentioned above. Algorithm Machine Learning ・Rust ・Python ・OpenCV ・PyTorch ・TorchServe ・Elasticsearch ・Vertex AI gRPC Infrastructure Google Cloud, Google Kubernetes Engine, Anthos Service Mesh, Istio, Cloudflare, Argo Workflows Event Bus Cloud Pub/Sub DevOps GitHub, GitHub Actions, ArgoCD, Kustomize, Helm, Terraform, Datadog, MixPanel, Sentry Key Points for Technology Selection While selecting technologies based on product characteristics, we also focus on standardization across teams to achieve overall optimization. Data Cloud SQL(PostgreSQL), AlloyDB, BigQuery, dbt, trocco To leverage the advantages of type systems in development, we prioritize the selection of static typed languages 01 02 GraphQL Dev Tools GitHub Copilot, Figma, Storybook Communication Slack, Discord, JIRA, Miro, Confluence Authentication Auth0
27 その他 SaaS Elastic Cloud Architecture Overview Drawing Files Cloud Storage Tenant/User DB Cloud SQL Drawing DB AlloyDB Drawing Analysis System Workflow Cluster GKE Multiple Instances Argo Workflows Cloudflare Workers Cloudflare Drawer BFF GKE Pod Drawing API GKE Pod Search API GKE Pod File API GKE Pod User API GKE Pod Drawer UI GKE Pod Drawing Analysis APIs Vertex AI Drawing Analysis APIs Vertex AI *For internal system integration, we also utilize Pub/Sub, Cloud Tasks, and similar technologies. shape similarity, symbol recognition, maximum dimension estimation, etc. PDF splitting, grayscale conversion, and various analytical processing calls, etc.
29 Global product × Local sales The products we are developing are for the global market, and actually established a sales office in Vietnam to sell. Can be involved not only in the development of Japanese products, but also in the development of products to be sold in our own country. Contributing to the future of Manufacturing Despite being a major global industry, manufacturing has seen little innovation in the procurement domain. Working at CADDi VN means tackling these challenges head-on and innovators in your home country's flagship industry. Development of in-house products CADDi Vietnam is not an offshore development base, everyone can take ownership of the development of our own products. In fact, CADDi Quote is a product that was jointly developed in Vietnam and Japan! The engineering organization in Vietnam has only about 12 engineers, and is still in the early stages of product and organizational development. Most new joiners will be at the forefront of development and involved in the start-up phase of building up the organization and the market. Development Team Members 1 2 3 4 5 6 Be Distinctive While it is still growing, the CADDi as a whole is successfully raising funds in the billions of yen, product sales are growing like T2D3 and are being used by well-known enterprise companies in Japan, such as major automobile manufacturers. That is why we can think big and jump into new challenges. CADDi's engineering team is composed of experts, including Kaggle Grandmasters and former CTOs. Their exceptional expertise fuels the organization's strength. The thrill of building a business in a chaotic startup environment A solid foundation for new challenges
of the Manufacturing Industry Technology is not an "end" but a "means." We select the necessary technologies for problem-solving as needed. Engineering is not about creating functions; it is about delivering true value to the user.
colleagues, and yourself. 4 Uphold Integrity Accomplish what cannot be achieved alone through the collective strength of the team. Let’s create the best CADDi together. 3 As One First, excel in your own capabilities. With your exceptional skills, dedicate yourself to fostering the growth of others 2 Be Distinctive Aim for ambitious objectives. Instead of pursuing one goal, strive for two or three. 1 Think Big
Drawing Data Intelligence Team Bao Tran (Bryan) Backend Engineer Vinh Kieu (Vinh) Engineering Manager Anh Trinh (Stephen) Frontend Engineer Hung Phan (Henry) Machine Learning Engineer Nghia Liu (Louis) Machine Learning Engineer Hiroki (Yamamon) Head of Engineering An Nguyen (An) Engineering Manager Drawer / Workflow Engine Team Tri Van (Jimmy) Backend Engineer Anh Nguyen (Anton) Backend Engineer Duy Le (Nolan) Backend Engineer Long Do (Long) Backend Engineer Members in Vietnam Technology Team Bach Nguyen (Bach) Backend Engineer
a career as a software engineer at GMO Internet, progressing to Tech Lead at Persol Technology. Joined CADDi in 2022, initially focusing on developing solutions in manufacturing technology and currently leading the application engineering team in the Vietnam office. Duy Le Workflow Engine Team Backend Engineer Tech startup enthusiast with experience building scalable platforms, data pipelines, and analytics. Proven track record in solving complex problems, driving innovation, and delivering impactful solutions in dynamic, fast-paced environments. Anh Nguyen Workflow Engine Team Backend Engineer He is a Backend Engineer with experience in developing a CCG gacha game for Gree company. After join CADDi, he is working on a critical platform project, building a workflow that enables Drawer to convert and analyze thousands of drawings within few minutes. Hung Phan Drawing Data Intelligence Team Machine Learning Engineer With a background in ML, he’s been engaged in research and development of machine vision and high-performance computing (HPC) systems since 2016 in Vietnam and Norway. After contributing to automotive software and manufacturing solutions at Bosch Vietnam, he joined CADDi in 2022 as a MLE, where he developed the drawing translation feature. Currently, he’s part of the Drawing Feature Extraction team, focusing on advancing Technical Drawing Understanding. Hiroki Yamamoto Head of Engineering Managing Vietnam local development team at GREE, Inc, he became the representative of the local center. Developing IoT devices at UMITRON, Inc after his coming back to Japan. Being attracted by the speed and scale of business, he joined CADDi in 2022 as an Engineering Manager and now he is the Head of Engineering in Vietnam Office, leading CADDi VIETNAM Technology Team. Members in Vietnam Technology Team 34 An Nguyen Engineering Manager Machine Learning Engineer She is a research scientist with a PhD in Computer Science from Keio University in September 2016. Previously, she worked as an AI Technical Solution Leader at Bosch. Since 2022, she has held the position of Engineering Manager at CADDi, where she oversees projects focused on drawing feature extraction solutions.
and test containers • Unit Testing Principles, Practices, and Patterns • CADDi Core UI • Chrome extension related stuff • Encouragement of telepresence: Let’s make it easier to develop software on k8s • Knowledge Hub with Vector database • Kustomize vs Helm (vs Terraform) • Are you still using chrome without AI? • Design Tokens • Understand GraphQL and DataLoader PERFECTLY • Tailscale, easiest way how to build VPN Previous Topics Team Building Hackathon Lightning Talk 37
‧Working hour: 8am-5pm (Flextime is available with the approval of Managers) ‧Many employees have kids, and the company values family time. ‧Hybrid ‧Employees can work both on-site and remotely ‧A recommended in-office day is Wednesday to enhance collaboration, where Scrum events and meetings are concentrated. This allows for focused development on other days ※Feel free to discuss working styles and office attendance during interviews. Hybrid Work Model Information flow and communication ‧Tech All Hands（Weekly online meetings） to share management messages and technical strategy. ‧Company-wide All Hands（Weekly online meetings） management messages, business updates, customer use cases, and product wins ‧Off-site Meetings ‧Technology Division Socials(Bi-weekly) casual meetups open for all employees. ‧About 20% of members are from countries outside Japan. ‧Company-wide All Hands and Tech All Hands are conducted in both Japanese and English. ‧Key announcements and documents are provided in both languages. ‧Some teams primarily conduct business in English. Bilingual Environment
System ‧Holidays : Based on Vietnam National Holidays, Year-end and New Year holidays ‧Transportation allowance : VND1,000,000 per month if coming to Office once per week ‧Child Allowance: VND1,000,000 per month per child ‧Marriage Leave and Bonus ‧Maternity Leave and Bonus ‧Parental and Care Leave ‧Equipment Provision: PCs and monitors provided. ‧Company Kick-Off and Parties: Held every six months (hybrid format). ‧Team Meal Support: once per month Team Cohesion Initiatives Health Management ・Annual Health Checkups ・Premium Health Insurance ・External Training Support ・Server Cost Subsidy Growth Support
40 Someone who loves to continuously face difficult challenges, yet constantly questions whether the challenge at hand is truly the most important, and what the real, most crucial challenge is. Facing challenges, but questioning them Of course, you love technology. However, you also care about the people surrounding the technology and products, striving to create something together by uniting "technology and users. Loves technology, but cares about people too Someone who enjoys thinking for themselves about what needs to be done, what they want to do, and how they want to be, and running towards it. Yet, they also love doing things with others that they can't accomplish alone, and sharing the joy of facing challenges with others. Loves running solo, but wants to achieve something with others A person who is always sharpening their own skills. However, when it comes to solving problems, they don't cling to the skills they have or are refining, and instead, consider what is the most optimal tool to solve the issue. Constantly honing their skills, but not fixated on them
caddi_eng
iamctodd
moore
gr2m
erikaheidi
csswizardry
sugarenia
paulrobertlloyd
addyosmani
dougneiner
reverentgeek
smashingmag
rocio
