Software Engineer - Production AI Workflows

How to apply

Email ***email_hidden*** with subject: Software Engineer - [Your Full Name]. Include your CV, GitHub/portfolio, and 2-3 examples of production systems you built or shipped. Applications that do not follow this format may not be reviewed.

About the role

We are building Maxify, an AI-native operating platform where agents own real business workflows, act through company systems and remain accountable to source data, permissions, evaluations and measurable outcomes.

We are looking for a Software Engineer to build and operate production AI workflows on top of our shared platform. You will translate validated business processes into reliable agents that connect to company data, tools and systems and produce measurable outcomes.

This is not a prompt-writing or chatbot role. You will own complete workflows from discovery and tool design through testing, deployment, monitoring and improvement.

What you will own

• Validate users, pain points, sources of truth, baseline metrics and expected value before building.
• Model business processes as explicit states, actions, decisions, approvals and exception paths.
• Build business-specific agents that complete production workflows end-to-end.
• Develop tools, structured outputs and context strategies.
• Integrate approved APIs, databases and enterprise systems.
• Use deterministic code where probabilistic model behavior is unnecessary or unsafe.
• Create evaluations for completion, accuracy, safety and business outcomes.
• Implement monitoring, escalation, exception handling and human-review paths.
• Diagnose failures and convert repeated issues into tests, guardrails or platform requirements.
• Measure revenue, cost, customer-experience or operational-attention improvement.

What we are looking for

• Strong production software engineering ability in Python and/or TypeScript.
• Experience building LLM applications, agents or substantial workflow automation.
• Experience with APIs, webhooks, databases and asynchronous processing.
• Ability to translate ambiguous operations into explicit system behavior.
• Understanding of tool calling, structured outputs, retrieval, context management and model limitations.
• Experience testing deterministic and non-deterministic systems.
• Strong debugging skills across instructions, application code, data, tools and external services.
• Good judgment about when to use an LLM, deterministic code or human approval.
• Ability to connect engineering work to measurable business outcomes.
• Clear written communication, high agency and rapid learning.

Strong signals

• You have shipped an agent or automation that performs real actions in production.
• You can demonstrate measurable business impact from something you built.
• You design approval, exception and escalation paths instead of assuming perfect model behavior.
• You use AI development tools heavily while verifying and owning the result.
• You can explain a failed workflow run, identify the root cause and show the prevention mechanism.
• You understand that reliable agents require source data, workflow state, tools, evaluations and observability, not prompts alone.

Nice to have

• Experience with OpenAI Agents SDK, LangGraph, OpenClaw or comparable systems.
• Experience with Lark/Feishu, Shopify, CRM, accounting or enterprise SaaS integrations.
• Experience with workflow automation or operations-heavy products.
• Familiarity with knowledge graphs, ontology design or governed memory.
• Experience working directly with business teams or customers.

What this role is not

• Prompt engineering without code or operational ownership.
• Building isolated chatbot demonstrations.
• Pure machine-learning research or model training.
• Pure frontend, data-science or infrastructure work.
• Allowing agents to act without evaluations, permissions or traceability.

Success in the first 30 days

• Validate one business workflow, its source of truth and baseline performance.
• Map its states, tools, decisions, approvals and exception paths.
• Ship at least one meaningful production improvement.
• Deliver one production workflow on the shared platform.
• Add evaluations covering normal, edge and failure cases.
• Implement monitoring, escalation and human-review paths.
• Demonstrate measurable improvement in speed, quality, cost or customer experience.
• Identify at least one reusable requirement for the shared platform.

These outcomes are subject to timely access and platform readiness.

How we work

• We move quickly and build for permanence.
• We start from the business outcome and source of truth.
• We use deterministic systems where they are more reliable.
• We treat evaluations, observability and recovery as part of the product.
• AI accelerates the work; it does not remove engineering accountability.

Interview process

Shortlisted candidates will complete an onsite prototyping test based on a practical AI workflow. We evaluate working output, engineering judgment, effective use of AI tools, debugging ability and clarity of explanation.