Taki Haruki

About

Principal Software Engineer and AI/ML Platform Architect with 10 years building full-stack applications and AI-powered platforms for insurance, pension, and workforce analytics. Bridges frontend development, distributed backend systems, and ML/LLM pipelines for regulated industries; directed teams up to 8 engineers to deliver enterprise platforms and measurable improvements in underwriting and document review efficiency. Currently architecting an enterprise AI platform across MLOps/LLMOps and model governance, including METI-aligned governance templates and developer enablement via Backstage.

Experience

• Principal Software Engineer · Sompo Holdings05-01-2025

  − Architecting enterprise AI platform incorporating MLOps, LLMOps, and model governance capabilities with projected 40% reduction in AI deployment timelines − Defining AI governance framework templates addressing model risk management and bias monitoring aligned with Japanese METI guidelines − Constructing developer experience layer using Backstage enabling self-service AI development with appropriate guardrails across the organization − Advising senior leadership on AI strategy, build versus buy decisions, and talent development priorities for WTW Japan's AI advisory practice − Mentoring 8 engineers on LLMOps best practices and prompt engineering techniques, establishing knowledge-sharing sessions adopted across 3 regional offices

• Senior Software Engineer · Sompo Holdings05-01-2023 – 04-30-2025

  − Led development of LLM-powered underwriting decision support system processing 200+ applications daily, cutting medical document review time from 45 to 15 minutes − Designed actuarial modeling platform integrating GPT-4 and Claude APIs, accelerating model documentation by 60% and shortening development cycles by 25% − Built LLMOps infrastructure including prompt versioning and output evaluation pipelines with LangSmith ensuring reproducibility and systematic improvement − Established guardrails and validation layers ensuring generated actuarial code met quality and compliance standards before human review − Coordinated cross-functional collaboration between actuarial, compliance, and engineering teams, completing platform delivery 3 weeks ahead of schedule

• AI Engineer · Marsh McLennan10-01-2020 – 02-28-2023

  − Engineered commercial insurance risk scoring platform on AWS with sub-500ms API response time, compressing initial risk assessment from 4 hours to 45 minutes − Architected end-to-end ML pipeline for Mercer workforce analytics under Japan APPI regulations, attaining 82% accuracy in employee attrition prediction − Created climate risk analytics platform integrating PostGIS with IPCC scenario projections, supporting TCFD-aligned disclosures covering 50,000+ properties − Managed technical teams of 6-10 engineers, conducting code reviews and ensuring delivery quality standards across concurrent engagements

• AI/ML Engineer · Sompo Holdings06-01-2018 – 10-01-2020

  − Developed gradient boosting mortality prediction models using XGBoost demonstrating 12% improvement in discriminatory power versus traditional actuarial life tables − Constructed pension risk analytics dashboard with PostgreSQL and daily refresh capabilities serving 15 corporate plan sponsors managing 500+ billion yen in combined assets − Implemented ML-based auto insurance pricing models with Python and FastAPI real-time rating engine processing 10,000+ daily quotes at 99.9% availability − Promoted to Senior AI Engineer following successful delivery of pricing optimization project during COVID-19 remote transition without delay

• Data Engineer · Brainpad Inc.07-01-2016 – 04-30-2018

  − Designed Kafka-based data ingestion layer for regional bank customer segmentation, processing 6 months of transaction history into 200+ behavioral features − Delivered production ETL pipelines processing 10+ million telecom customer records daily with 99.5% uptime, shrinking data preparation time from weeks to days − Built automated model retraining pipeline with monthly cluster refresh and CRM integration, driving campaigns with 23% higher response rates − Conducted exploration analysis on 50+ million POS transaction records, identifying 12 high-potential cross-selling opportunities for major retail chain

• Junior Backend Engineer · Works Applications09-01-2015 – 06-30-2016

  − Developed backend API endpoints for COMPANY HR module serving 40% of large Japanese enterprises, trimming batch processing time by 15% − Implemented RESTful APIs for HUE next-generation ERP platform using Spring Framework with sub-200ms response time benchmarks

Education

• The University of TokyoBachelor’s Degree, Computer Science

Similar talent on Pangea

EErese R.Senior Full-Stack & AI Engineer | 10 Years Experience (AI/LLMs, RAG, Full-Stack)Hong KongNNigel B.Senior AI Software Engineer | LLM/RAG, Distributed Systems & Cloud-Native Platforms (6+ Years)Castelo Branco, PortugalChetan M.I help founders prototype AI features fast enough to test demand in 2–3 weeks | Senior AI & Software Engineer | ex–Google Cloud CR MentorMontreal, QCKerry D.Senior AI/Fullstack Engineer | 7+ Years in Fintech & Banking etc PlatformsEverett, MA 02149