Pursuing research in episodic memory and multi-agent systems. Exploring algorithms for dynamic neural oscillations role in memory function.
Improved search quality on Uniqlo EC store search platform. Worked on ranking algorithms and search infrastructure for enterprise-scale e-commerce. Details confidential.
Optimized API response times for toB product by 96% (3 minutes → 7 seconds) through Rails query optimization. Maintained production environment in service while refactoring critical paths for enterprise customers.
Built commercial-grade web application from scratch in two weeks. Led a team of 4 engineers to deployment while maintaining clean architecture. Designed and deployed infrastructure using AWS CDK.
Proposed AWS cost optimization for map-based data platform. Analyzed resource utilization and software architecture to identify inefficiencies. Recommended ¥1.2M/month cost reduction.
Implemented automated E2E testing for quantum/HPC PaaS platform using Playwright. Reduced test execution time from 3 hours to 5 minutes. Created documentation and scripts to enable future maintainability.
Full-stack development on AWS. Phase 1: IaC with Terraform (100+ AWS resources), Lambda batch processing, GitHub Actions CI/CD. Phase 2: Analytics on BigQuery, image recognition debugging for edge devices.
Coursework in basic mathematics, physics in mechanical engineering, and computer science. Focused on machine learning, deep learning, and neural information processing.
Proficient in full-stack web development and ML systems. Most experienced in infrastructure concerns.
Experience shipping production systems. From prototyping to continuous integration and deployment.
Architect cloud systems at enterprise scale. Experience with IaC, CI/CD, containerization, and cost optimization across major cloud providers.
Training and fine-tuning large language models. Experience with distributed training, search optimization, and data analysis pipelines.
Emphasis on clean architecture, rigorous communication, and reproducible workflows. Strong technical writing and documentation skills.
Harnessing and orchestrating LLM agents to write code, debug, and optimize systems. Balancing velocity and reliability in an era of AI-assisted development.
Led model training team for open-source LLM fine-tuning on challenging HLE benchmark. Achieved the best performance through SFT + DPO on Qwen3-235B. Managed distributed training across 8 nodes (H100s), coordinated 10-person cross-functional team, and made critical execution decisions under time pressure.
Built decision-support tool for enterprise prediction markets using LLM agents. Designed system where LLM agents participate as expert forecasters, using LangChain for prompt engineering and Google Vertex AI for inference.
Reduced API latency by 96% through Rails query optimization. Analyzed database access patterns, eliminated N+1 queries, and restructured data loading. Delivered 3-minute → 7-second response times for enterprise customers without downtime.
Analyzed multi-client AWS infrastructure to identify cost reduction opportunities. Built scripts for resource auditing and automated log retention. Identified ¥1.2M/month savings potential through instance right-sizing and redundant resource elimination.
Automated production E2E testing for quantum/HPC PaaS platform using Playwright. Created comprehensive test suite reducing manual testing from 3+ hours to 5 minutes. Documented workflow and created reusable patterns for team.
Get in touch if you want to work on a project with me. Open to roles in web infrastructure, LLM systems, or technical team leadership.