Original listing text, shown exactly as published by the company.
Your Mission
As a Senior Agentic Engineer at Hawk, your mission is to architect and build the agentic core of our AI Investigation Agent — a new product that fundamentally rethinks how human analysts collaborate with autonomous AI to investigate financial crime. You will own the hardest problems in the system: how the agent reasons, what context it sees, what tools it has, how we measure whether it's getting better or worse, and how we keep it reliable when it's making consequential decisions about real financial crime cases. You'll set the architectural direction for how Hawk builds agentic products, and you'll write the code that proves it works.
Key Responsibilities
- Architect and build the agentic systems that power the Investigation Agent: workflow design in LangGraph, tool design and integration, multi-step reasoning, memory, and the orchestration patterns that hold it all together.
- Own context engineering as a first-class discipline. Decide what the agent sees, when it sees it, and how it's structured. Most agent failures are context failures — your job is to make sure ours aren't.
- Build the feedback loops, eval harnesses, and tracing infrastructure (Langfuse and beyond) that turn agent development from vibes into a measurable, iterative engineering discipline. You make the invisible visible.
- Make consequential architectural decisions: how we structure agent state, how we handle long-running workflows, how we recover from failure, how we deploy and roll back agent changes safely, how we keep latency and cost in check at scale.
- Own complex backend engineering work end-to-end across our stack: Python, FastAPI, Postgres, Redis, Kubernetes. Agentic systems are still software systems — they need to be reliable, observable, and operable like any other production service.
- Partner closely with Product/UX, Data Science, and the rest of Engineering to shape what gets built. You bring strong opinions about what's actually possible with current models and frameworks, and you push back when a proposed feature is the wrong shape for an agent.
- Champion an AI-native engineering workflow. You use Claude Code, Cursor, or equivalent agentic coding tools as a daily part of how you work, and you help raise the bar for how the team uses them. If you find yourself doing the same thing twice, your natural instinct is to automate it.
- Help establish the foundational engineering practices for agentic products at Hawk: how we test agents, how we evaluate them, how we deploy them, how we monitor them, and how we know when to trust them.
Your Profile
- 5+ years of professional software engineering experience, with deep backend expertise in Python and a strong track record of shipping production systems that other people depend on.
- Hands-on experience designing and building agentic systems — multi-step workflows, tool use, context management, memory, evaluation. You've done this in LangGraph, in another framework, or from scratch, and you have opinions about which approach fits which problem.
- Strong intuitions about context engineering, prompt design, eval design, and the failure modes of LLM-based systems. You understand the difference between a demo that works once and a system that works reliably at scale.
- Solid backend engineering fundamentals: FastAPI (or equivalent), Postgres, Redis, Kubernetes, observability, and the operational discipline to run all of it in production.
- Hands-on fluency with agentic coding tools (Claude Code, Cursor, or equivalent) as part of your daily workflow. You have an opinion about what they're good at, what they're not, and how to get the most out of them.
- Excellent holistic thinking and problem-solving skills. You shape the product and the architecture rather than execute pre-defined tickets.
- Excellent communicator with a collaborative mindset, capable of articulating technical trade-offs to both technical and non-technical stakeholders.
Bonus
- Production experience with Langfuse or equivalent LLM observability and eval tooling.
- Prior experience in Anti-Financial Crime (AML, fraud, compliance, or investigations tooling) — or a deep curiosity to go learn the domain fast.
- You have shipped agentic products to real users in production — not just prototypes, demos, or internal tools. You've felt the difference between "the agent worked when I tried it" and "the agent works for our customers, every day, at scale."
