Hiroshi K.
Senior Data Scientist
Experience
Oct 2024 - Sep 2025
1 yearMunich, Germany
Senior Data Scientist
Siemens
- Addressed inconsistent annotation quality across distributed teams by implementing a tiered review system with structured feedback templates, reducing rework by 35% and improving inter-annotator agreement to 0.85 Cohen's kappa.
- Solved factual accuracy degradation in AI-generated technical documentation by developing a multi-stage verification workflow combining automated fact-checking with expert human review, achieving 98% accuracy on manufacturing specification documents.
- Eliminated sensitive content leakage in training data by designing PII detection filters and establishing redaction protocols for IoT sensor data and maintenance logs, ensuring compliance with GDPR and internal security policies.
- Streamlined annotation team onboarding by creating comprehensive training workshops and mentorship programs, reducing ramp-up time from 6 weeks to 3 weeks while maintaining quality standards.
- Resolved guideline interpretation inconsistencies by developing detailed annotation protocols with concrete examples and edge cases, decreasing clarification requests by 60% across multilingual annotation projects.
- Fixed quality drift in ongoing annotation projects by implementing statistical process control charts and automated quality checks, catching deviations 3 days earlier than manual review processes.
- Improved annotation throughput without sacrificing quality by optimizing review workflows and implementing batch processing of similar content types, increasing daily output by 25% while maintaining 95%+ accuracy.
- Addressed feedback loop inefficiencies by establishing structured peer review sessions and weekly calibration meetings, ensuring consistent application of annotation guidelines across all team members.
Jun 2021 - Sep 2024
3 years 4 monthsWarsaw, Poland
Data Scientist
EPAM Systems
- Solved financial document classification inconsistencies by developing a hierarchical annotation system with clear decision trees, improving classifier F1-score from 0.78 to 0.91 on banking transaction data from SAP and internal databases.
- Addressed annotation scalability challenges for multi-language financial reports by implementing a distributed labeling platform with quality gates, processing 50K+ documents monthly with 94% consistency across English and German content.
- Fixed model performance degradation in production by establishing continuous annotation pipelines for hard cases, reducing false positives in fraud detection by 22% while maintaining 99.9% recall on transaction monitoring systems.
- Resolved training data quality issues by implementing schema validation and outlier detection in feature pipelines, decreasing data-related production incidents by 65% across client financial services applications.
- Eliminated annotation backlog during project scaling by designing efficient batch processing workflows and priority queuing systems, maintaining SLA compliance despite 3x volume increases during quarterly reporting periods.
- Improved cross-team annotation consistency by conducting weekly calibration sessions and developing detailed feedback documentation, achieving 0.88 inter-annotator agreement across distributed teams in different time zones.
- Solved model interpretability challenges in client presentations by creating comprehensive annotation guidelines with business context, reducing explanation time from 45 to 15 minutes during stakeholder reviews.
Aug 2019 - Jun 2021
1 year 11 monthsWarsaw, Poland
Senior MLOps
Fujitsu
- Addressed manual annotation bottlenecks in manufacturing defect detection by implementing semi-automated labeling tools with human verification, reducing labeling time by 40% for CV models processing production line imagery.
- Solved training data versioning chaos by establishing centralized annotation storage with metadata tracking, enabling reproducible model training across multiple manufacturing facility datasets.
- Fixed quality inconsistencies in sensor data annotation by developing standardized labeling protocols and conducting train-the-trainer sessions, improving model accuracy by 15% on predictive maintenance tasks.
- Resolved annotation tool reliability issues by migrating from custom scripts to containerized labeling applications, achieving 99.5% uptime for distributed annotation teams across three manufacturing sites.
Apr 2014 - Aug 2019
5 years 5 monthsJapan
Machine Learning Engineer
Fujitsu
- Solved initial data labeling challenges for early ML projects by developing structured annotation guidelines and quality check procedures, establishing foundation for reproducible model development.
- Addressed limited training data availability by implementing data augmentation techniques and systematic labeling workflows, enabling successful deployment of first-generation recommendation systems.
- Fixed annotation consistency issues across team members by creating detailed examples and edge case documentation, improving model performance stability during initial production deployments.
- Resolved manual quality assurance bottlenecks by developing automated validation scripts for annotated datasets, reducing review time by 50% while maintaining high data quality standards.
Summary
10+ years of experience building and deploying machine learning systems across manufacturing and financial services domains. Specialized in end-to-end MLOps implementation with expertise in data annotation quality assurance, factual accuracy evaluation, and content moderation workflows. Proven track record in establishing quality control processes for AI training data and implementing structured feedback systems for annotation teams. Combines deep technical ML expertise with practical experience in mentoring junior staff and conducting quality assurance reviews.
Languages
Japanese
NativeGerman
AdvancedEnglish
AdvancedEducation
Apr 2016 - Mar 2018
Tokyo University of Science
Master's of Computer Science · Computer Science · Japan
Apr 2010 - Mar 2014
Tokyo University of Science
Bachelor's of Computer Science · Computer Science · Japan
Certifications & licenses
Google Professional Machine Learning Engineer
AWS Certified Machine Learning - Specialty
AWS
DeepLearning.AI Natural Language Processing Specialization
DeepLearning.AI
Need a freelancer? Find your match in seconds.
Try FRATCH GPT More actions
Similar Freelancers
Discover other experts with similar qualifications and experience
- English
- Deutsch
2025 © FRATCH.IO GmbH. All rights reserved.