Senior AI Engineer – Conversational Agents & Workflow Orchestration

Full Description

We are looking for a Senior AI Engineer to take ownership of an existing production conversational agent used to manage operational workflows (mail handling, forwarding, scanning, compliance, etc.).

This is not a greenfield project.

The agent already exists and is deployed.

The role is to understand, stabilize, improve, and extend the current system, including logic design, conversational flows,and execution reliability.

You will work closely with the founder, who has already defined much of the business logic, gate requirements, and intent taxonomy.

Your role is to turn that into a robust, scalable, production-ready agent.

What You’ll Be Responsible For

Conversational Agent Ownership

• Take over an existing LLM-powered agent built with FastAPI + OpenAI (GPT)

• Fully understand the current architecture, datasets, gate logic, and failure cases

• Own the agent end-to-end: intent → gate → path → execution → logging

Agent Logic & Orchestration

• Implement and maintain gate-based workflows (state machine style)

• Correctly handle alternative satisfaction paths (Path A / B / C… are alternatives, not steps)

• Implement dynamic routing / branching when the next step depends on user intent

• Ensure the agent always asks the correct minimal follow-up question

• Prevent repeated questions, meta-prompt leakage, and logic loops

Prompt & Dataset Engineering

• Improve and maintain:

• system / master prompts

• judge prompts

• follow-up (ask) prompts

• Work with structured datasets (Excel/CSV) defining:

• intents

• gate order

• satisfaction paths

• follow-up templates

• Keep datasets and code perfectly aligned

Reliability & Debugging

• Analyze agent logs and trace failures

• Fix cases where the agent:

• selects the wrong gate

• ignores provided information

• asks off-topic or repeated questions

• Add guardrails and validation to prevent regressions

Collaboration & Ownership

• Proactively suggest improvements (not just implement tickets)

• Clearly explain agent behavior and decisions in plain English

• Think like an agent architect, not just a coder

Required Skills & Experience

Must-Have

• Strong experience with Python

• Experience building LLM-based conversational agents

• Hands-on with OpenAI / GPT models

• Experience with FastAPI or similar backend frameworks

• Understanding of state machines / workflow orchestration

• Strong reasoning about ambiguous user input and intent resolution

• Ability to design deterministic logic around probabilistic models

Very Important

• You understand that:

• “Paths” are alternative ways to satisfy a gate

• Gate order is not always static

• The agent must reason based on meaning, not keywords

• You are comfortable debugging logic failures, not just syntax errors

Nice to Have

• LangChain or similar frameworks (not required)

• Experience with production AI systems

• Experience deploying on Render, AWS, or similar platforms

• Familiarity with vector stores (Chroma, Pinecone, etc.)

Job Types: Full-time, Part-time

Expected hours: 20 – 40 per week

Application Question(s):

• How many years of hands-on experience do you have building or maintaining production conversational AI or LLM-based systems (beyond simple demos)?

• How many years of experience do you have working with complex backend architectures (multi-step workflows, state machines, orchestration logic, or similar)?

• Have you personally owned, maintained, or significantly refactored a production AI system already in use by real users? Provide brief description

Work Location: Remote