Applied AI & retrieval

Source-grounded AI workflows with explicit review and escalation.

NextEleven’s applied-AI work includes implemented confidential client workflows and inspectable public retrieval software. The focus is evidence-backed retrieval, controlled tool use, understandable failure behavior, and clear separation between generated content and external actions.

Discuss an opportunity ← Back home

01Implemented capabilities

What the work demonstrates

The focus is useful implementation and inspectable behavior—not generic AI promises or business outcomes I cannot verify.

Retrieval

Answers with sources

Multi-source ingestion, FTS5 and vector retrieval, reciprocal-rank fusion, reranking, citations, and retrieval evaluation.

Backend

Python service interfaces

FastAPI, SSE, REST APIs, workflow orchestration, provider integration, automated tests, and operational documentation.

Mobile

Kotlin and Jetpack Compose

Project experience connecting Android interfaces and multi-device services to Python application and retrieval workflows.

Tools

Explicit actions

LLM tool calling, MCP, structured inputs, provider routing, validation, and clear separation between generated text and external actions.

Safety

Human review and escalation

Workflows that pause for approval, expose uncertainty, and escalate instead of silently inventing an answer or action.

Privacy

Local and confidential boundaries

Self-hosted retrieval options and public descriptions that withhold private client, product, health, source, and operational details.

02Evidence boundaries

What I am—and am not—claiming

Implemented confidential work.A multi-component applied-AI system was implemented, demonstrated, and documented as client work; current production adoption is not claimed.
Inspectable public work.PersonalRAGVault is MIT licensed, and Harness exposes public proof of its agent and tool architecture while remaining proprietary.
No outcome claim.No clinical, financial, operational, or business outcome is claimed without specific evidence and an explanation of its limits.
No confidential identities.The client, private product, and sensitive subject matter remain unnamed.

Discuss the system or the engineering behind it.

For a technical role, implementation engagement, or focused engineering discussion, contact Sean directly and identify the project that brought you here.

View résumé Email Sean