The Role
Design, develop, and deploy production-grade agentic AI and multi-agent systems for enterprise workflows. Build tool integrations, MCP servers, memory and retrieval systems, evaluation suites, safety guardrails, human-in-the-loop controls, observability, and reliable containerized services. Integrate with enterprise platforms and APIs while controlling agent cost, latency, and failure modes. Create reusable templates and client-ready demonstrations in collaboration with AI solution leaders.
Summary Generated by Built In
Key Responsibilities
A. Agent Design & Development
• Design and build AI agents that plan and execute multi-step tasks using tool calling, structured outputs,
and state management.
• Build multi-agent systems — orchestrator/worker, planner/executor, and review patterns — and decide
when a single agent is the better choice.
• Translate business processes from BA and solution documents into agent workflows with clear goals,
steps, decision points, and exit conditions.
• Design agent memory and context: conversation state, long-running task state, and retrieval of relevant
knowledge at each step.
B. Tools, Integrations & MCP
• Build and maintain the tools agents use — API connectors, database queries, document actions, and
browser or desktop automation where no API exists.
• Develop Model Context Protocol (MCP) servers that expose client systems to agents in a secure, reusable
way.
• Integrate agents with enterprise platforms such as Salesforce, ServiceNow, SAP, Microsoft 365, Google
Workspace, and telephony systems.
• Write clear tool descriptions, input schemas, and error messages so agents use tools correctly and recover
when a call fails.
C. Reliability, Safety & Evaluation
• Build evaluation suites for agent behaviour: task success rate, tool-selection accuracy, step count, cost per
task, and failure analysis on real traces.
• Design human-in-the-loop checkpoints for high-impact actions — approvals, escalations, and handoff to
human agents.
• Enforce guardrails: least-privilege tool access, prompt injection defences, action limits, timeouts, and safe
fallbacks.
• Instrument full traceability — log every model call, tool call, and decision so any agent run can be replayed
and audited.
• Control cost and latency through model routing, caching, context management, and limits on runaway
loops.
D. Deployment & Collaboration
• Deploy agents as reliable services — containerised, observable, with queueing and retry handling for longrunning tasks.
• Prepare demo-ready agent workflows for client pitches, working with the AI Solutions Lead on scenarios
that show clear business value.
• Contribute reusable agent templates, tool libraries, and MCP connectors to Flatworld's AI accelerator
library.
• Document agent architecture, tool inventories, permissions, and known limitations for every build.
Requirements
Mandatory Technical Requirements
The following are non-negotiable for this role:
• Python: Strong production-grade Python — clean, typed, tested code; async programming; API
integration. [MANDATORY]
• LLM Tool Calling: Hands-on experience building applications with LLM tool/function calling and structured
outputs on at least one major API (Anthropic, OpenAI, Google, or Azure OpenAI). [MANDATORY]
• Agent Frameworks: Built at least one working agent with LangGraph, CrewAI, AutoGen, OpenAI Agents
SDK, Claude Agent SDK, or a well-structured custom orchestration loop. [MANDATORY]
• System Integration: Experience integrating with REST APIs, webhooks, and authentication schemes
(OAuth, API keys, service accounts). [MANDATORY]
• Agent Evaluation & Tracing: Practice of testing agent behaviour with scenario-based test sets and
inspecting execution traces — not just manual trial runs. [MANDATORY]
• Deployment: Ability to containerise with Docker and deploy services to a cloud platform; proficiency with
Git. [MANDATORY]
Strongly Preferred
• Model Context Protocol (MCP): Building or deploying MCP servers and clients.
• RAG: Retrieval pipelines and vector databases (Pinecone, Qdrant, Chroma, or pgvector) used as agent
knowledge sources.
• Observability: Agent tracing and evaluation tools such as LangSmith, Langfuse, Arize Phoenix, or
OpenTelemetry.
• Workflow & Queues: Durable workflow or task-queue tools (Temporal, Celery, Redis, or cloud-native
equivalents) for long-running agent tasks.
• Cloud Agent Platforms: AWS Bedrock Agents, Azure AI Foundry Agent Service, or Google Vertex AI Agent
Builder.
Advantageous
Not required, but a clear differentiator for this role:
• Voice agents — real-time speech-to-text, text-to-speech, and telephony integration for contact centre
automation.
• Browser or computer-use agents (Playwright, Selenium, or model-native computer use) for systems
without APIs.
• Background in RPA (UiPath, Automation Anywhere, Power Automate) and moving rule-based bots to
Agentic workflows.
• Experience with enterprise platforms such as SAP, IBM Maximo, or ServiceNow.
• Awareness of data protection requirements (GDPR, HIPAA, India's DPDP Act) as they apply to autonomous
systems.
Benefits
What We Offer
• Work at the front of applied AI — building agents that automate real enterprise processes, not just chat
interfaces.
• Variety across industries and systems, with a direct line from your build to a client decision.
• Access to current commercial and open-weight models and freedom to choose the right framework for
each problem.
• Mentorship from the AI Solutions Lead and AI Solution Architect, with exposure to pre-sales and solution
design.
• Learning budget for AI upskilling, conferences, and cloud certifications.
• Competitive compensation with a clear path toward Senior Agentic AI Engineer or AI Solution Architect
tracks
Skills Required
- Strong production-grade Python skills, including clean, typed, tested code, asynchronous programming, and API integration
- Hands-on experience building applications with LLM tool or function calling and structured outputs using Anthropic, OpenAI, Google, or Azure OpenAI APIs
- Experience building at least one working agent with LangGraph, CrewAI, AutoGen, OpenAI Agents SDK, Claude Agent SDK, or a well-structured custom orchestration loop
- Experience integrating REST APIs, webhooks, and authentication schemes such as OAuth, API keys, or service accounts
- Experience testing agent behavior with scenario-based test sets and inspecting execution traces
- Ability to containerize services with Docker and deploy them to a cloud platform
- Proficiency with Git
- Experience building or deploying Model Context Protocol servers and clients
- Experience with retrieval pipelines and vector databases such as Pinecone, Qdrant, Chroma, or pgvector
- Experience with agent tracing and evaluation tools such as LangSmith, Langfuse, Arize Phoenix, or OpenTelemetry
- Experience with durable workflow or task-queue tools such as Temporal, Celery, Redis, or cloud-native equivalents
- Experience with AWS Bedrock Agents, Azure AI Foundry Agent Service, or Google Vertex AI Agent Builder
- Experience with voice agents, speech-to-text, text-to-speech, or telephony integration
- Experience with browser or computer-use agents using Playwright, Selenium, or model-native computer use
- Background in RPA tools such as UiPath, Automation Anywhere, or Power Automate
- Experience with enterprise platforms such as SAP, IBM Maximo, or ServiceNow
- Awareness of GDPR, HIPAA, or India's DPDP Act as applied to autonomous systems
Am I A Good Fit?
Get Personalized Job Insights.
Our AI-powered fit analysis compares your resume with a job listing so you know if your skills & experience align.
Success! Refresh the page to see how your skills align with this role.
The Company
What We Do
Flatworld Solutions is a global IT, business consulting, and outsourcing firm founded in 2004. As a diversified BPO and technology services provider, it supports over 18,000 clients across 100 countries. The company leverages automation and Generative AI to help enterprises streamline operations, enhance efficiency, and achieve transformative growth through a wide range of professional services and IT solutions.
