Senior Forward Deployed Engineer

Praxent•Remote - CO, FL, GA, MA, MD, MN, NE, NC, OR, PA, SC, TN, TX, WA, WA

2d•$205,000 - $250,000•Remote

About The Position

Praxent builds software for financial services companies. We are an Anthropic partner, and building agentic AI systems inside our clients' environments is now a core part of the work. This role exists because that work needs a different kind of engineer than a traditional project team supplies — someone who sits with the client, decides what should get built, and builds it. This is the first Forward Deployed Engineer hire at Praxent. You would be defining the role, not inheriting it. This role has been categorized as a Remote position. Applicants for this position must be currently and legally authorized to work in the United States without the need for current or future sponsorship. Praxent will not offer immigration sponsorship or assume sponsorship of an employment visa for this position. International relocation or remote work arrangements outside of the U.S. will not be considered.

Requirements

6+ years shipping production software, full-stack. Python is what most of this work is written in; .NET or Node are fine substitutes. React or Next.js on the front end.
Experience leading the technical side of a delivery. You have owned an implementation plan, sequenced the work, and directed other engineers without being their manager.
API design and systems integration. You have integrated against something old, poorly documented, and load-bearing, and made it hold.
Real client-facing delivery experience. You have run a discovery session, handled a difficult stakeholder, and delivered an account without someone above you managing the relationship.
Production experience with LLMs and agents. Prompt engineering, agent development, evaluation frameworks, deployment at scale. Application engineering, not ML research.
Comfort with ambiguity. You deliver value before all the requirements are defined, because they never are.
Range in the room. You are precise with engineers and clear with executives, and you can switch between those two rooms inside an hour.
A story about scoping down. A time you told a client not to build something, and why you were right.
Willingness to travel to client sites 20–30% of the time.

Responsibilities

Own technical delivery end to end on your accounts — discovery, architecture, build, production, and the messy weeks after launch.
Embed with client engineering and business teams. Translate a VP's business problem into an architecture, and that architecture back into language the VP can defend to their board.
Build production applications and agentic systems inside client environments: LLM-backed workflows, agent orchestration, evaluation frameworks, and integrations against core banking, lending, policy administration, and financial data systems.
Design the integration surface. Most of this work is API design and systems integration against platforms that were not built to be integrated with.
Harden prototypes into systems a regulated institution will actually run — security review, audit trail, PII handling, performance under real load. The distance between a working demo and a production deployment at a bank is most of the job.
Scope down. Identify when a client is asking for the wrong thing and talk them into the right thing. This is the part of the job we care most about.
Guide the client's own engineers. Part of leaving well is that their team can extend what you built without calling us.
Own the hardest production problems on your accounts — the ones that surface at 4pm on a Friday in someone else's infrastructure.
Own the implementation plan and direct the work. You decide the sequence, and as an engagement scales you delegate to the engineers around you — including the internal AI engineers who will work for you. This is a technical leadership role, not a people-management one.
Make the case for growth by building things worth expanding. You carry no revenue target and you do not run the commercial conversation — our Client Partners do that. Your argument is the solution itself.
Flag risk before it becomes a problem.
Leave behind documented, runnable, maintainable software that still works after you rotate off.