Seiro Iwamoto岩元清郎
Turns ambiguous needs into specifications and acceptance criteria, then owns application-layer implementation and verification. Go / Python, AWS / Azure / GCP. Over ten years in backend engineering.
Discuss a November project →- Bring an AI/LLM PoC to a deliverable state with authentication, operations, and acceptance testing
- Turn ambiguous needs into technical requirements and acceptance criteria, then implement them
- Trace root causes across application, data, retrieval, generation, and cloud layers
- Engagements requiring regular on-site work
WORK
6 CASE STUDIESAI/LLM delivery — from requirements to acceptance
Translates ambiguous business needs and environment constraints into technical requirements and acceptance criteria, then owns delivery across frontend, API, RAG, infrastructure, and verification. Closes the gap between a PoC and a testable delivery environment while keeping answer provenance visible to users.
OPEN SHEET →Delivering a multimodal document-processing PoC
Moved a document-processing PoC combining an image-capable VLM API with PDF vector analysis into an authenticated service that a customer could validate in its cloud environment. Locked down inference routing, identity, secrets and service operations, then isolated a real-data regression and an LLM quota failure by layer.
OPEN SHEET →Taking a research PoC into the product — knowledge graph analysis in a multi-tenant product
Ported a knowledge graph analysis and visualization feature, validated by a research team on a dedicated server, into a product where many customers share the same platform. Rebuilt it around the product's UI structure, data model, and permissions. When it ran out of memory in the staging environment before production, added stopgaps that protect other users — a cap that rejects oversized inputs before heavy computation, and serialized heavy work — and split the root fix into a separate, scoped issue.
OPEN SHEET →Big data SaaS platform
GKE Manual→Autopilot migration, a MySQL→Datastream→BigQuery data pipeline, and construction and operation of Airflow pipelines, handled continuously over two years.
OPEN SHEET →Tech Lead in an English-language environment
Evaluated the issues in existing GCP infrastructure and redesigned it from scratch starting at the VPC with Terraform. Set up CI/CD, selected the security-assessment vendor, and ran scrum and hiring in English.
OPEN SHEET →Consumer web app designed from scratch
Designed the AWS architecture for a consumer web app from scratch (API Gateway / VPC / ECS / CloudFront / WAF). Set up SSO with OpenID Connect and CI/CD. Handled it as tech lead.
OPEN SHEET →- FIG.1AI/LLM delivery — from requirements to acceptance
- FIG.2Delivering a multimodal document-processing PoC
- FIG.3Taking a research PoC into the product — knowledge graph analysis in a multi-tenant product
- FIG.4Big data SaaS platform
- FIG.5Tech Lead in an English-language environment
- FIG.6Consumer web app designed from scratch
SKILLS
ABOUT
The FDE title did not mark a sudden change in how I work. I had long owned backend and cloud technology choices through implementation and operations; more recently, that responsibility extended upstream into customer discovery, technical requirements and acceptance criteria, and downstream into validating AI/LLM application delivery. I use Palantir’s definition of the FDE role as market vocabulary for that working model, not as a claim of affiliation or product experience. I work fully remotely from Fukuoka, using written specifications, tests, UI checks, and real-data evidence in the same delivery cycle.