Distinguished Software Engineer

Credit Acceptance•Washington, DC

1d•$232,500 - $341,000•Hybrid

About The Position

Credit Acceptance is proud to be an award-winning company recognized both locally and nationally across multiple workplace categories. Our world-class culture is shaped by dedicated team members who are driven to succeed as professionals individually and together as a team. Backed by a strong product, exceptional people, and a stable financial foundation, we’ve grown into a leading provider of used and new car financing across the country. Our Engineering and Analytics Team Members utilize the latest technology to develop, monitor, and maintain complex practices that help optimize our success. Our Team Members value being challenged, are encouraged to express their ideas, and have the flexibility to enjoy work life balance. We build intrinsic value by partnering with all functions of our business to support their success and make strategic business decisions. We focus on professional development and continuous improvement while enjoying a casual work environment and Great Place to Work culture! Credit Acceptance makes vehicle ownership possible for consumers other lenders decline, through a dealer network, an originations and servicing platform, and pricing and decisioning models refined over decades. The technology estate underneath this business was built project by project over a long period, and modernizing it is the largest engineering opportunity at the company. As the Distinguished Engineer, you are the senior-most individual contributor in Technology, reporting directly to the Chief Technology Officer. You will set the multi-year technical strategy for the company while staying close enough to the code and systems to remain credible — this is not an architect role that reviews designs from a distance, nor a strong engineer role that goes deep in a single domain. Two things sit at the center of the job: the target architecture for the enterprise, and the architecture for running AI agents across it. You will not own a team; you will own the hardest technical problems the company has, and get them resolved by convincing the people who do own the teams. This position will work from home and also require some work onsite at an assigned office work location.

Requirements

Bachelor’s degree in Computer Science, Information Systems, or closely related field of study; or equivalent work experience.
Minimum 10 years of software engineering experience or comparable depth of experience.
A track record of setting and delivering multi-year technical strategy at enterprise scale, as an individual contributor rather than through a reporting line.
Depth in distributed systems architecture, and in at least two of: agentic or LLM-based systems in production; large-scale cloud migration of transactional systems; modern data platform and lakehouse architecture; machine learning platform and model serving; high-traffic consumer or marketplace applications.
Architecture ownership of agentic or LLM-based systems that reached production, including how tool and data access were governed, how the system was evaluated, and what its failure modes turned out to be.
Demonstrated ownership of a legacy-to-modern migration of systems the business runs on, including the sequencing, the cost case, and the failure modes, with production revenue at stake rather than only familiarity with the concept.
Hands-on credibility today: actively reads code, writes proofs of concept, joins design reviews, and participates in incident analysis.
Evidence of influence without authority across organizational boundaries, including work that changed direction at the executive level.
Strong written communication; architecture here is carried by documents that other people can act on without the author in the room.
Demonstrated ability to develop staff and principal engineers, and to raise the technical bar of engineers not directly managed.
Experience delivering in a regulated industry, or a clear ability to work inside audit, security, and compliance constraints without treating them as an obstacle.

Nice To Haves

Recognized contribution beyond one employer, such as open source, standards work, publications, patents, or a sustained external technical presence.
Consumer lending, auto finance, insurance, or payments experience; or large-scale distributed systems experience supporting a business-critical function in another high-stakes domain.
The Model Context Protocol or comparable agent tool-integration standards, and identity and authorization design for non-human actors.
AWS and Databricks at production scale.
Experience standing up an experimentation platform, or moving an organization from tens to thousands of experiments per year.
Experience building AI-native development practice across a large engineering organization, and measuring whether it worked.
Marketplace, search, or inventory systems with more than one class of seller.

Responsibilities

Convene the principal and staff engineers who own each domain to define the end-state architecture across originations, servicing, marketplace, data, and AI, and use it to drive the multi-year plan. Ensure every component has a defined interface with stated inputs and outputs, every unmade decision has a written principle standing in for it, and the document states honestly how close the organization is to the end state. Publish the architecture, including the tradeoffs that were rejected and why.
Own the enterprise agent architecture spanning servicing and collections, the dealer side, originations, and internal engineering and back-office workflows, built on one common foundation rather than one stack per team. Establish a governed way for agents to discover and call business capabilities, data, and tools, using the Model Context Protocol as the integration layer. Define an identity and authorization model that treats an agent as a first-class actor with its own least-privilege entitlements, along with observability, evaluation, and cost controls that work consistently for every agent. Set policy on what an agent may decide independently versus what stays with a person, and define where that line moves and on what evidence. Design the architecture to produce an examiner-ready record of what each agent saw, recommended, and touched, and set the evaluation and adversarial testing approach that makes agent outcomes defensible. Ensure the same foundation and controls apply to agents built by outside firms on the company's behalf.
Own the technical strategy for migrating transactional and analytical workloads off on-premises infrastructure to the company's strategic targets — AWS for transactional workloads and Databricks for analytical workloads — determining what lifts and shifts, what replatforms, what gets re-architected on the way versus after, and what the steady-state cost looks like. Own the resilience and recovery architecture for the target state.
Set the architecture for change-data-capture-based ingestion, streaming where warranted, and a modern warehouse with no job-scheduling dependency, moving the organization from largely batch analytical data movement to real time. Help settle the open question of who owns the transactional and analytical data models.
Architect for pricing and credit model and scorecard updates that drop in as a layer rather than requiring a multi-quarter rebuild, and for an experimentation platform that supports orders of magnitude more concurrent experiments than a conventional release process allows.
Own the architecture and interfaces for a unified vehicle inventory, consumer search, and lead routing platform, ensuring the data model supports more than one class of seller from the start.
Define the golden path for how software is specified, generated, reviewed, tested, and deployed, including the analyzers and gates that enforce it, and measure the inputs to that practice rather than just the outputs. Hold the build-versus-buy bar for AI capability, building what is core competency and about to scale, and renting the rest.
Own the instrumentation, baseline, and remediation architecture needed to meet the standard that no page takes more than one second. Own peak-event headroom, including understanding the limits of each system and maintaining tested degradation levers ahead of need.
Move the organization from project-shaped systems toward owned platforms with stated contracts. Document every component's inputs and outputs, and be ready to argue why closely coupled components should not be consolidated into one. Work toward a two-week requirement-to-production iteration bar, treating architecture that cannot be delivered incrementally as unfinished. Name what comes off the plan whenever a change in direction is requested. Make the call and own the outcome on architectural recommendations. Treat security, privacy, auditability, and regulatory obligation as design inputs from the start in this regulated environment.
Stay personally engaged in the technical work rather than directing it from a distance. Join design reviews, write proofs of concept, read code, and engage in incident analysis. This hands-on time does not replace the work the teams own, and it is not a fallback when delivery slips. Every engineer in the organization will read how you write, how you argue a design, how you handle being wrong, and how you treat engineers who disagree with you as the bar. Staff and principal engineers should be measurably better at this work for having worked with you.
A published target architecture the domain owners agree with and that the annual and long-range plans are actually written against; an enterprise agent foundation that internal teams and partners build on, with access controls, evaluation, and audit records in place; a credible, dated, costed migration path off on-premises infrastructure and demonstrated progress against it; measured improvement in chosen input metrics such as latency, availability, recovery objectives, deployment frequency, and experiment throughput; growth in the number of principal and staff engineers who are better at this work as a result of this role; and architectural decisions that stop being relitigated because the reasoning was written down and the interfaces were clear.