Lead Software Engineer

Nationwide Property & Appraisal Services LLC•Franklin Township, NJ

14d•Remote

About The Position

We are looking for a lead engineer to work alongside the founding engineer from a recent acquisition. The role is split between hands-on engineering and technical oversight of other engineers (today: occasional contractors and AI-assisted work; likely to grow). You will own features end-to-end — from task brief through implementation, testing, documentation, and shipping — and you will review the work of other engineers (human and AI-augmented). You will work within a specification-driven methodology that uses AI-augmented development (Claude in planning, Claude Code in execution) under direct human review. This is not a “ship code fast” role. Our core product is in a regulated domain where mis-described features can create compliance risk. The pace is deliberate: write the brief, review the architecture, execute, review, ship, document.

Requirements

Solid Django and Django REST Framework experience. You can spin up a viewset, write a serializer, and reason about query performance without reaching for documentation for the basics.
Comfortable working full-stack: JavaScript (vanilla; no specific framework required), HTMX, and reading and modifying Django templates with embedded HTMX attributes. You do not need to be a frontend specialist, but you need to be able to extend an existing HTMX screen, write the JavaScript glue when needed, and reason about the request/response cycle from browser through Django and back.
Comfortable with Celery, PostgreSQL, Redis, and Docker Compose for local development.
Experience with — or willingness to learn quickly — the regression and statistical analysis libraries in the codebase (pandas, numpy, statsmodels, scikit-learn). You do not need to be a statistician, but you need to be able to read and modify analytical code responsibly.
Has shipped real code using Claude Code, Cursor, Copilot, or similar — not just experimented. You know the failure modes, you do not blindly trust AI output, and you do not dismiss it either.
Experience reviewing other engineers’ work — formally as a tech lead, informally as a senior on a team, or via open-source maintainership. You can give feedback that is specific, kind, and uncompromising about the things that actually matter.
Strong writing skills. You can write a clear task brief, a useful lesson entry, a non-misleading commit message, and a PR review comment that helps rather than alienates.
A temperament for working in a regulated domain.
Must have the ability to associate and work well with others including co-workers, vendors, suppliers and customers.
Must have 5-7 years’ experience in current or related field.
Must have strong computer skills.
Be able to work in a fast-paced environment.
Being able to communicate effectively, good listening skills, deal with difficult calls, problem solve.
Must be able to multitask.

Nice To Haves

College Degree preferred.
General knowledge of processing, underwriting or appraisals a plus.

Responsibilities

Implement features in Django (5.x) and Django REST Framework against written task briefs.
Work on both the backend analytical engine and the HTMX frontend. Our core product is a single codebase: when a feature needs both a REST endpoint and an HTMX screen to consume it, you will build both halves.
Maintain strict API parity between the HTMX frontend and the REST API — every capability available in one must be available in the other, returning identical data shapes.
Use Claude Code (in VS Code) as your execution partner: writing scoped prompts, reviewing AI-produced code, catching AI mistakes before they ship.
Review pull requests from other contributors. Catch decoupling and compliance violations before they land.
Help less-experienced engineers structure their work within the methodology — writing briefs, scoping prompts, reviewing diffs, maintaining the session-management docs.
Mentor on AI-assisted development practices specifically: how to write Claude Code prompts that produce reliable output, when to trust AI suggestions and when to push back, how to verify AI output rather than rubber-stamp it.
Maintain the architectural mandates as living rules, not slogans. When someone proposes a shortcut that would violate decoupling or compliance framing, your job is to redirect.
Keep documentation in sync as part of every feature: PROJECT_SNAPSHOT.md, CLAUDE.md, OPEN_TICKETS.md, and LESSONS_BANKED.md.
Bank lessons when something surprises you or another engineer. Add tickets when you discover deferred work.
Hold the line on the two non-negotiable architectural mandates: Decoupling: backend owns data and business logic; no HTMX-only code paths; the service layer is the shared backbone. Compliance: regulatory disclosure text in core/disclosures.py requires sign-off from our regulatory reviewer before any changes; product framing follows specific language rules.