WORK IN JAPAN: Senior Visual AI Research Engineer in Japan (HIRING NOW)
Dreams and waking.
Theory and practice.
Research and development.
Each of these are two sides of the same coin.
This role is an opportunity to deliver the full coin.
Welcome to our guide on becoming a Senior Visual AI Research Engineer in Japan.
This position is open right now.
There’s a lot to cover, so let’s get started.
What is this role?
AI is moving quickly, but some of the most valuable AI opportunities are not just about building general-purpose models.
They are about applying advanced AI to deep, complex, industry-specific problems.
That is exactly what this company is doing in manufacturing.
This rapidly growing manufacturing AI company in Japan develops SaaS products that help manufacturers eliminate inefficiencies and make better decisions by leveraging industrial data, including CAD drawings, 3D models, specifications, quotations, documents, defect records, and other multimodal manufacturing information.
They are now hiring a Senior Research Engineer / Senior Research Scientist focused on 2D/3D vision, multimodal understanding, and applied AI research.
This is a strong opportunity for someone who wants to work at the intersection of advanced research, real-world industrial data, and production impact.
Role Overview
This role focuses on researching, developing, and implementing AI systems that can understand diverse manufacturing data, especially:
2D technical drawings
3D CAD data
Specifications and manufacturing documents
Defect reports
Geometry, annotations, symbols, dimensions, and tolerances
Multimodal combinations of visual, textual, and structural data
The goal is not research for research’s sake.
While academic contributions such as publications are encouraged, the main focus is to turn research outcomes into practical business value.
You will work on research initiatives over a 6–12 month horizon, from defining the research scope to building prototypes, conducting PoCs, validating results with real-world data, and helping translate those results into product or business impact.
Why This Position Matters
Manufacturing is one of the largest industries in the world, but much of its most important information is locked inside highly specialized data formats.
General-purpose AI models are powerful, but they are not always designed to understand complex industrial data such as CAD drawings, geometric structures, tolerances, annotations, manufacturing symbols, or defect records. These require deeper domain-specific understanding.
This role exists to help the company build a long-term technical advantage in manufacturing AI.
By developing AI systems that can reason across manufacturing data, you will help improve critical decision-making processes such as:
Design reviews
Quality assessments
Defect risk prediction
Cost optimization
Similar drawing search
Missing dimension detection
Layout and symbol interpretation
In other words, this is a role where research can directly influence how real manufacturing decisions are made.
Core Responsibilities
1. 2D / 3D Geometry Understanding
A major part of the role involves developing AI methods that can understand complex geometric structures specific to manufacturing.
This may include extracting meaningful local features from 3D data, connecting those features with semantic information, and building models that align 2D drawings with 3D representations.
Examples of the kinds of problems you may work on include:
Understanding geometric structures in manufacturing drawings
Interpreting dimensions, tolerances, and annotations
Connecting 2D drawing information with 3D CAD data
Extracting useful features from 3D models
Building AI systems that can reason about manufacturing-specific geometry
This is especially relevant for candidates with experience in computer vision, 3D vision, CAD, B-Rep, GD&T, geometric processing, or related fields.
2. Multimodal AI and Vision-Language Model Enhancement
Another key area is multimodal AI.
You will work on models that integrate different types of manufacturing data, such as drawings, CAD data, specifications, and defect reports. The role also involves improving vision-language models for manufacturing-specific tasks.
General-purpose VLMs are often biased toward text and may struggle with technical drawings, industrial diagrams, or manufacturing-specific visual information. This role focuses on overcoming those limitations and developing models that perform well in this specialized domain.
Potential work includes:
Developing multimodal models for drawings, CAD data, and text
Improving VLMs for manufacturing use cases
Creating or extending benchmarks for domain-specific evaluation
Improving models for similar drawing search
Detecting missing dimensions
Interpreting layouts, symbols, and technical drawing information
Building systems that can reason across visual, textual, and geometric inputs
This makes the role a strong fit for someone interested in applied multimodal AI beyond standard image-text datasets.
3. Research to Business Impact
A defining feature of this position is the expectation that research should lead to measurable real-world value.
You will lead PoC initiatives using actual production data and collaborate with Product Managers, Engineers, ML teams, data platform teams, and domain experts. The work should ultimately contribute to better decision quality, improved operational efficiency, and stronger product capabilities.
Examples of applied impact areas include:
Defect risk prediction
Cost optimization insights
Manufacturing process improvement
Decision-support systems
Production-ready AI capabilities
Research prototypes that can move toward product implementation
This is a good role for someone who enjoys research but also wants to see their work used in real products.
Required Skills and Experience
The company is looking for someone with deep expertise in at least one of the following areas:
Machine learning
Computer vision, especially 2D or 3D vision
NLP
Multimodal AI
You should be comfortable reading, understanding, and implementing state-of-the-art research. Strong programming ability is also important, especially in Python, as the role involves experimentation, evaluation, and API implementation.
You should also have experience working with practical engineering environments, including:
GPU environments
Docker / containerization
Git / CI/CD
Cloud platforms
ML experimentation and evaluation
Real-world ML problem definition
Model development and continuous improvement
Because this role sits between research, engineering, product, and domain expertise, communication skills are also important. You need to be able to explain technical decisions clearly, collaborate across teams, and work with people who may come from different professional backgrounds.
Language Requirements
Because this role involves collaboration with Product Managers, Engineers, ML teams, data platform teams, and domain experts, strong Japanese language skills are important.
You should be ready to explain complex research ideas clearly and work effectively across different functions.
Our recommendation is at least JLPT N2 level Japanese.
Preferred Skills
You do not need to have every preferred skill, but the following experience would help you stand out:
3D CAD
B-Rep
GD&T
Geometric processing
OCR
Document understanding
Vision-language models
MLOps
Production ML pipelines
Distributed systems
GPU optimization
ML project or team leadership
Publications at top conferences
Patents
Open-source contributions
This role is especially interesting for someone who has strong research depth but also wants to apply that expertise to complex, messy, high-value industrial data.
Tech Stack
The broader technical environment includes:
Backend: Python, Rust, TypeScript
Frontend: TypeScript, React, Next.js
ML: PyTorch
Infrastructure: Google Cloud, GKE, BigQuery
APIs: GraphQL, REST, gRPC
Tools: Docker, Terraform, GitHub Actions
Monitoring: Datadog, Sentry
For this specific role, Python, PyTorch, GPU environments, Docker, cloud platforms, and ML experimentation workflows are especially relevant.
Nice-to-Have Experience
Experience with AI-related data infrastructure would be a strong advantage.
This includes experience or interest in:
RAG
Context-aware AI systems
Agent orchestration
Embeddings
Knowledge graphs
Vector databases
Data catalogs
Semantic layers
Metadata management
AI-ready data infrastructure
You do not need to be an ML Engineer, but you should be interested in how data platforms can support the next generation of AI-powered products.
Because part of the role involves making data more discoverable and usable by AI agents, experience thinking about metadata, structured data access, and query accuracy would be highly relevant.
Who This Role Is Best For
This role is likely a strong fit for you if you are an AI researcher, research engineer, machine learning engineer, computer vision engineer, or applied scientist who wants to work on difficult, domain-specific AI problems.
It may be especially attractive if you:
Want to work on multimodal AI with real industrial data
Are interested in 2D / 3D vision and geometry understanding
Enjoy bridging research and production
Want ownership over research initiatives
Prefer applied research with measurable business impact
Are excited by ambiguous, open-ended technical problems
Want to build AI systems that solve problems general-purpose models cannot easily handle
This is not a narrow model-tuning role. It requires someone who can define problems, explore research directions, build prototypes, validate impact, and work with product and engineering teams to move promising ideas closer to real-world use.
Compensation and Work Style
The role is based in Tokyo with a hybrid work style.
Compensation is listed at approximately:
¥12M – ¥20M annual salary
Additional compensation details include:
Salary review twice per year
Stock options available
Hiring Process
The typical hiring process takes around one month and includes:
Optional casual interview
Resume screening
Technical assignment
HR discussion
Technical interviews
Final interview
Offer
The technical assignment and interviews will likely be important opportunities to show not only your research knowledge, but also how you think through real-world ML problems, evaluate trade-offs, and connect technical work to business impact.
How to Stand Out as an Applicant
To stand out for this role, you should show evidence that you can work across both research and implementation.
Strong candidates will likely be able to demonstrate:
Deep knowledge in ML, computer vision, NLP, or multimodal AI
Ability to understand and implement recent research papers
Experience applying ML to real-world problems, not only academic benchmarks
Strong Python and PyTorch skills
Comfort working with GPUs, Docker, cloud platforms, and modern engineering workflows
Clear thinking around evaluation, benchmarking, and model improvement
Interest in manufacturing, CAD, geometry, or industrial data
Ability to communicate research clearly to non-research stakeholders
If you have publications, patents, open-source projects, production ML experience, or examples of applied research projects, those would be highly relevant.
For this role, it is not enough to say “I have worked with AI.” You will want to show what kinds of models you built, what problem they solved, how you evaluated performance, what trade-offs you considered, and how the work created impact.
Career Path and Future Opportunities
This role can lead in several directions depending on your strengths and interests.
You could grow as a senior individual contributor specializing in multimodal AI, computer vision, 3D understanding, or domain-specific industrial AI. You could also move toward research leadership, ML engineering leadership, or product-facing AI strategy.
Possible future paths include:
Principal Research Engineer
Applied AI Research Lead
Multimodal AI Specialist
Computer Vision / 3D Vision Lead
ML Platform or MLOps leadership
AI Product Strategy
Research-to-product leadership
Because the company is building AI capabilities around highly specialized manufacturing data, this role gives you the chance to develop expertise in an area that may become increasingly valuable as companies look for AI systems that go beyond generic models.
FAQ
Is this more of a research role or an engineering role?
It is both. The role involves research, experimentation, and prototype development, but the main goal is to create business value through real-world implementation. Candidates who only want academic research with no product connection may find the role too applied. Candidates who enjoy turning research into usable systems will likely find it a strong fit.
Do I need manufacturing experience?
Manufacturing experience is helpful but not necessarily required. The JD emphasizes deep expertise in ML, computer vision, NLP, or multimodal AI. However, experience with CAD, B-Rep, GD&T, geometric processing, OCR, document understanding, or VLMs would be a strong advantage.
What kind of AI problems will I work on?
You may work on 2D / 3D geometry understanding, multimodal AI, VLM enhancement, similar drawing search, missing dimension detection, layout and symbol interpretation, defect risk prediction, and cost optimization insights.
Is this role suitable for someone from academia?
Yes, especially if you are interested in applied research and can implement your ideas in practical systems. Publications are valued, but the primary goal is translating research into business impact.
What kind of candidate is the company looking for?
The company is looking for someone with strong technical depth, ownership, curiosity, communication skills, and the ability to balance research excellence with real-world impact. You should be comfortable working in an ambiguous and fast-changing environment.
