~/
$ whoami

takuya-ogata

// Applied Research Engineer

// Bringing AI products from research to production.

about

Data Science Consultant at Accenture Japan (2021— · promoted 2023.12).

End-to-end ownership: requirements → modeling → full-stack build → deploy → ops.

Lead teams in LangChain · LangGraph · RAG · multi-agent territory.

Building and running pairkan.com as a personal product on the side.

Looking for Applied Research roles where I can engage from the “what to build” stage.

now

  • - Data Science Consultant @ Accenture Japan
  • - Building pairkan.com (Next.js 16 + Supabase + Cloudflare Workers, solo)
  • - Working on a multi-agent LLM workflow with the strategy team (LangGraph · Next.js · FastAPI · Azure AI Foundry)

experience

  • 2021.06 — present Accenture Japan Ltd.
  • └── 2023.12 Promoted to Data Science Consultant
  • ─
  • 2019 — 2021 M.S. Computer Science Kyushu University
  • 2015 — 2019 B.Eng. EECS Kyushu University
  • ─
  • 2024 AWS Certified Solutions Architect — Associate

selected_work

[01] pairkan.com personal · 2025— → /en/pairkan/
  • Consumer web service on Next.js 16 + Supabase + Cloudflare Workers.
  • Solo full-stack — requirements through UI, DB schema, auth, and deploy.
Next.js 16 React 19 Supabase Cloudflare Workers Tailwind v4 shadcn/ui
[02] multi-agent-assessment work · 2025.12 — 2026.01
  • Multi-agent LLM workflow for structured decision support.
  • Inputs are reviewed against domain literature to produce actionable hypotheses & follow-up questions.
  • Co-built with strategy team; owned AI workflow + UI/back-end + job server.
LangGraph Next.js FastAPI AWS Azure AI Foundry
[03] realtime-3d-agent work · 2025.02 — 2025.04
  • CXO-grade 3D-character realtime response system.
  • Owned requirements → pipeline → DevOps & coaching.
OpenAI Realtime API WebRTC
[04] customer-support-rag lead work · 2024.10 — 2025.02
  • ReAct + RAG agent for counter-service automation, full-stack.
React Flask LangChain RAG OpenAI
[05] speaking-test-scoring lead work · 2022.12 — 2023.08
  • ASR + scoring pipeline for automated speaking-test grading.
  • Whisper segmentation → BERT scoring · GPU parallel.
PyTorch Whisper BERT

archive

click to expand · 9 items
  • · multiagent-marketing work · 2025.05 — 2025.07 AWS Neptune · SVO graph
  • · supply-chain-pmo work · 2025.09 — 2025.11 PMO · roadmap
  • · accounting-ai lead work · 2023.12 — 2024.04 pandas · luigi · LightGBM · ChatGPT
  • · multi-genai-2023 work · 2023.09 — 2023.12 QA-manual app · BPR support
  • · rag-feasibility work · 2023.08 — 2023.09 OpenAI Embedding · ChatGPT
  • · pharmacy-churn work · 2023.06 — 2023.09 Python · scikit-learn · KPI design
  • · web-nav-analysis work · 2022.06 — 2022.10 PySpark · GA · cost optimization
  • · text-analytics-poc work · 2022.03 — 2022.06 NLP · annotation design
  • · ma-tool-dev work · 2021.08 — 2022.02 Django · React · SageMaker · Step Functions

github

$ gh api users/kyouyap | jq
public_repos
37
total_stars
1
followers
10
fetched_at
2026-05-11
$ gh api repos | aggregate-languages
  • Python 45.1%
  • TypeScript 17.3%
  • Astro 9.9%
  • JavaScript 8.5%
  • R 7.5%
  • Shell 6.3%

skills

languages
Python · TypeScript · SQL · Shell
genai
LangChain · LangGraph · RAG · OpenAI / Anthropic API · Realtime API · multi-agent
ml
PyTorch · BERT · Whisper · scikit-learn · LightGBM · pandas
backend
FastAPI · Flask · Hono · Django · Quart
frontend
React · Next.js · Astro · Vue · Tailwind
data
PySpark · Databricks · AWS Neptune · networkx · luigi
infra
AWS · Azure AI Foundry · Supabase · Docker · PostgreSQL · Step Functions · SageMaker · Jenkins
tools
Vim · VSCode · Jupyter · Claude Code

contact

$ ./connect
  • github → @kyouyap
  • twitter → @kyouyap
  • email →