AI Solution Architect

Long Finch Technologies

United States

Posted On: Nov 19, 2026

Posted On: Nov 19, 2026

Job Overview

Experience

12 - 25 Years

Salary

Depends on Experience

Work Arrangement

Remote

Travel Requirement

0%

Required Skills

  • AI
  • LLM
  • RAG
  • MCP
  • DIGITAL
  • Python
Job Description

 

Job Description

SmartAssist is enterprise agent platform. It dynamically constructs specialized AI assistants at runtime from a governed catalog of skills and MCP tools, and serves them through the SmartAssist Agent Framework (Python / FastAPI). We are engaging a contract Specialized AI Agent Developer to design, build, and refine the specialized agents that run inside SmartAssist.

Each specialized agent is assembled from a skill definition retrieved via the Skills Registry and is restricted to the tools and procedures that skill authorizes. This role centers on agent behavior: authoring agent and coordinator definitions, engineering system and instruction prompts, shaping multi-step reasoning and reliable tool use, and ensuring agents operate correctly within the platform's role-based access, Open Policy Agent (OPA) enforcement, and strict token-budget constraints.

This is a hands-on engineering role rather than a research role. Success is measured by production-ready specialized agents that integrate cleanly into SmartAssist and behave predictably at scale.

 

Core Job Functions

·         Design and implement specialized agents in the SmartAssist Agent Framework, dynamically constructed from Skills Registry definitions and restricted to each skill's allowed tools.

·         Author and maintain agent definitions (e.g., .agent.md) and coordinator / orchestration prompts that route a user's selected use case to the correct specialized agent.

·         Engineer, test, and iterate on system and instruction prompts to produce reliable, repeatable tool use and multi-step task completion.

·         Integrate agents with platform services: the Skills Registry MCP (registry_get_skill, registry_get_tools_for_skill), the SkillsProvider, the APIM gateway, and OPA policy checks.

·         Optimize agent context toward the platform's working-context target (~4K tokens) while preserving accuracy across a catalog of 1,000+ skills and ~1,500 tools.

·         Validate agent behavior with the integration test suite and structured evaluations; diagnose and correct failure modes such as wrong tool selection, fabricated steps, or policy denials.

·         Collaborate with the MCP tool and skill development team to shape the tool and skill contracts agents depend on, and feed back improvements to skill design.

·         Uphold platform guardrails in every agent delivered: prompt-injection resistance, least-privilege tool scope, and full auditability.

·         Create evaluation suites and benchmark scenarios that validate skill execution and agent behavior.

·         Align with existing observability and logging capabilities to support agent execution tracing, tool-call analysis, and operational monitoring.

·         Create technical documentation, architecture diagrams, and other knowledge-transfer sufficient for internal teams to maintain delivered components independently.

Experience

·         Strong Python engineering experience, including FastAPI or a comparable async service framework.

·         Demonstrated experience building LLM-powered agents or assistants with an agent framework (e.g., Microsoft Agent Framework, LangChain, or similar) using function / tool calling.

·         Practical prompt engineering: system-prompt design, instruction tuning, and driving structured, reliable tool use from a model.

·         Working knowledge of the Model Context Protocol (MCP) and tool-calling patterns; able to reason about tool schemas and tool results.

·         Experience integrating enterprise LLM backends (e.g., Azure OpenAI and / or AWS Bedrock) and API gateways / auth (OAuth2, Entra, and JWT).

·         Comfortable with containerized delivery (Docker) and Git-based workflows.

·         Able to design evaluations and debug non-deterministic agent behavior methodically.

·         Nice to have: policy-as-code (OPA / Rego), role-based access control, multi-agent / coordinator orchestration, and experience in regulated or large enterprise environments.

 

 

 

 


Job ID: LFT122904


Posted By

Onima Kakkar

Recruiter