Profile

• Name: Daiki Chiba (千葉 大稀)

I am from Shiranuka, Hokkaido. After graduating from Hokkaido Kushiro Koryo High School, I studied physics at Niigata University and earned my bachelor’s degree in 2019. Since then, I have worked as a machine learning engineer across model research and development, MLOps, production operations, and dataset quality improvement. My primary focus is Computer Vision and Document AI, and I work across modeling, software, and infrastructure to solve real-world problems.

• Role: Machine Learning Engineer
• Kaggle: Competitions Expert (0 gold, 2 silver, 3 bronze medals)
• Contact: daikichiba.tech.ml[at]gmail.com

Focus & Interests

• Applied Machine Learning
  • Computer Vision and Document AI
  • Dataset design, annotation quality, and evaluation design
• ML Systems
  • MLOps and production operations
  • Data engineering and data management
• Current Areas of Interest
  • Search and information retrieval
  • Practical applications of multimodal AI, including vision-language models (VLMs)
  • Real-time inference for computer vision models
• Sustainable Product and Systems Development
  • Product development and systems architecture that deliver long-term value to users and the business
  • Evolvable, testable design that reduces maintenance and operational costs
  • Systems that make essential complexity explicit while minimizing accidental complexity
  • Development environments and systems where humans and AI agents can collaborate effectively

Links

• GitHub
• LinkedIn
• Wantedly
• Lapras
• Kaggle
• Zenn
• Bluesky
• SpeakerDeck

Work Experience

• July 2019–December 2021: Machine Learning Engineer

  • Developed a recommendation model
  • Researched and developed a predictive model for memory-device test data
  • Developed models for assessing English learners and built MLOps workflows with AWS and CircleCI

• January 2022–Present: Machine Learning Engineer

  • Own AI APIs for a document-processing SaaS end to end, from product requirements and R&D to production operations and maintenance
  • Improve document image understanding by revisiting not only model architectures but also annotation guidelines and dataset quality
  • Design and operate systems that sustain ML value over time, including inference cost optimization, training-data pipelines, HITL evaluation and continual learning, and document search

Competition Results

Kaggle

• Image Matching Challenge 2022 — Solo, 99th (Competition page)
• Child Mind Institute - Detect Sleep States — Team, 34th (Competition page)
• LEAP - Atmospheric Physics using AI (ClimSim) — Solo, 55th (Competition page)
• CZII - CryoET Object Identification — Team, 54th (Competition page)
• Yale/UNC-CH - Geophysical Waveform Inversion — Team, 18th (Competition page)

Community Competition

• 26-shinnen-3Dpathology — Team, 5th (Competition page)

atmaCup

• atmaCup #17 — Solo, 12th (Competition page)

License

Articles and code snippets on this blog are available under the following licenses:

• Articles: CC BY 4.0 — You may reuse them with attribution.
• Code snippets: MIT License — You may use them freely.

© 2024 Daiki Chiba