Lead I - ML Engineering

Role Description We are seeking a high-output Software Engineer to build production-grade AI agents, agent connectors, and supporting infrastructure for enterprise workflows. This role is focused on agent systems engineering , not pure model training. The work centers on building the body around the model: tools, connectors, memory, context management, orchestration, evaluation, and controlled execution. Current agent stacks increasingly depend on tool use, session/state handling, tracing, and sandboxed execution rather than model tuning alone.

A critical requirement for this role is that the engineer must also be AI-augmented in how they work . We are not looking for someone who merely builds AI systems in a traditional way. We want an engineer who already uses AI to perform 80-90% of the software engineering workflow

• requirements shaping, design, code generation, debugging, testing, documentation, and iteration while applying engineering judgment to validate, correct, and productionize the output. This role is about speed, leverage, and practical delivery.

Role Objective Design and deliver agentic systems that can perform meaningful work with minimal human intervention, while enabling the engineer to use modern AI tooling to compress end-to-end engineering cycle time. The engineer will build agents that can reason over context, use tools, call enterprise systems, persist state, learn from prior interactions, and operate safely inside governed cloud environments. Modern agent implementations now commonly rely on handoffs, tools, traces, sessions, and controlled sandboxes to do exactly this.

Key Responsibilities

• Design and build AI agents that automate meaningful end-to-end workflows

• Engineer agent body components including:

• enterprise connectors and APIs

• tool calling layers

• persistent memory and session/state handling

• context management, retrieval, and skills/pattern libraries

• orchestration across single-agent and multi-agent flows

• evaluation, tracing, and guardrails

• Build connectors to enterprise platforms, data stores, documents, and operational systems using standardized tool and protocol patterns where appropriate, including MCP-style integrations. MCP is an open protocol for exposing tools, resources, and prompts to LLM applications.

• Create robust cloud-native services to support agent execution, logging, observability, access control, and controlled deployments.

• Implement persistent and short-term memory patterns so agents can retain useful context without becoming noisy or brittle. OpenAI s current guidance explicitly highlights sessions, trimming, and compression as important context-engineering patterns for long-running agents.

• Build and refine workflows for agent handoffs, tool selection, safe action execution, and human-in-the-loop approval where required. Both OpenAI and Anthropic now describe simple, composable agent workflows and tool-based systems as the practical path to effective agents.

• Prototype quickly, test aggressively, and move the strongest agent patterns into stable internal use.

• Use AI directly in your own engineering workflow to accelerate requirements interpretation, system design, coding, testing, debugging, documentation, and iteration.

Required Qualifications

• Strong experience in software engineering, with a track record of building production-quality systems.
• Demonstrated hands-on experience building AI agents or agentic applications with tools, workflows, orchestration, or autonomous execution patterns.
• Strong experience with Python and/or TypeScript for backend or agent-system development.
• Experience with cloud infrastructure and modern deployment patterns across one or more major cloud platforms.
• Experience building connectors to APIs, databases, file systems, enterprise tools, or SaaS products.
• Experience with memory/state handling, context engineering, retrieval, or knowledge-grounded agent behavior. Persistent context and memory management are now treated as core implementation concerns in production agent systems.
• Experience with observability, debugging, tracing, or evaluation of complex application behavior. Agent tracing is now a formal capability in current agent SDKs.
• Strong judgment in separating what should be handled by prompts, retrieval, deterministic code, memory, or workflow design.

Critical Capability: AI-Augmented Engineering The successful candidate must already operate as an AI-augmented engineer . We expect the engineer to:

• use AI for the majority of day-to-day engineering work
• move from concept to prototype rapidly using AI-assisted design and implementation
• validate, correct, and productionize AI-generated output rather than handcrafting everything from scratch
• treat AI as a force multiplier in coding, testing, debugging, and documentation

This is not optional. The role is explicitly intended for someone whose working style reflects the future operating model we are trying to build.

Preferred Qualifications

• Experience with OpenAI Agents SDK or similar agent frameworks, including tools, handoffs, tracing, and sandboxed execution. OpenAI s latest SDK capabilities include file inspection, command execution, code editing, and controlled sandbox workspaces.
• Familiarity with MCP or equivalent agent/tool interoperability standards.
• Experience with multi-agent or workflow-based patterns, evaluator-optimizer flows, or modular agent architectures. Anthropic s current guidance recommends simple, composable agent patterns over overly abstract frameworks.
• Experience with secure execution environments, sandboxing, and human-in-the-loop safeguards. MCP guidance also emphasizes visible tool invocation and the ability for humans to deny tool actions.
• Exposure to ML engineering is helpful, but this is not primarily a model-training position.

What Success Looks Like

• Agent prototypes can be built in days, not months.
• Connectors and tools are reliable enough that agents can do useful work, not just generate text.
• Memory and context are engineered so agents stay coherent over long-running workflows.
• The engineer personally demonstrates exceptional productivity by using AI as the primary development accelerator.
• Internal users begin to see real autonomous or semi-autonomous outcomes, not just demos.

Skills ai agent,software engineering,cloud infrastructure,database,