Principal Software Engineer, AI

Pepper Auditors PLLC
Dallas, TX, United States
2 months ago
Apply on www.indeed.com
Prepare application

Role details

Contract type
Permanent contract
Employment type
Full-time (> 32 hours)
Compensation
$200,000.0 - $300,000.0
Working hours
Regular working hours
Job source

Tech stack

Artificial Intelligence Large Language Models Front End Software Development Virtual Agents

Job description

Pepper’s mission is to protect the integrity of every American’s employee benefit plan. We are a fast-growing team of technologists and CPAs working to build a modern audit firm that can serve as the fulfillment engine for federally mandated compliance work streams., We’re looking for an experienced software engineer to help build and scale our audit workbench; a pipeline-based harness - not a generic agent loop - that takes an engagement from raw evidence to reviewable workpaper. As of today that looks roughly like:

  • Ingestion and extraction
  • Recomputation and exception detection
  • Workpaper generation
  • Eval infrastrcuture
  • Frontend review surface

As one of the first engineering hires, you will help set the technical direction to ensure our audit wokrbench is usable and scalable from our our first 100 clients to our next 10,000.

The hard part

Most “build an AI agent” roles optimize for demo velocity. This one optimizes for a number a CPA stakes their license on. The reliability of this system comes from schema design, deterministic rule logic, validation, and structured human review - not from hoping the model is good. You’ll spend as much time on the deterministic-versus-stochastic boundary, eval harnesses, and traceability as on prompts and orchestration. The model does extraction and narrative drafting; the math and the regulatory logic are exact code. If “build something auditable, reproducible, and provably correct in a regulated domain” is more interesting to you than “ship a chatbot,” you’ll like it here.

Requirements

  • Have shipped production LLM systems where correctness mattered - extraction pipelines, agentic workflows, or document-understanding systems - ideally in a domain with real consequences (fintech, legal, healthcare, regtech).
  • Build evals as a reflex. You’ve maintained golden datasets and caught regressions before they reached users, and you treat “we can’t measure whether this got better” as an unacceptable state.
  • Have strong judgment about where to use a model and where to write a deterministic function - with a hard bias toward the latter when dollars or compliance are on the line.
  • Are a genuine 0 1 builder - you can own a system end-to-end, ship it, and operate it in production. You’re energized, not unsettled, by being the second engineer in the building.
  • Treat engineering hygiene as default, not overhead.

Benefits & conditions

Pulled from the full job description

  • 401(k), Nice to haves
  • Familiarity with accounting, audit, ERISA, 401(k) administration, or financial document processing. You do not need audit experience. Our partner CPAs own the domain end-to-end, and the regulatory logic lives in versioned, deterministic code you’ll help build.
  • Layout-aware / OCR extraction, retrieval systems, or structured-output generation.
  • Experience building internal tools or review-and-approval UIs.

Logistics

  • Compensation: Estimated $200,000-$300,000 base + equity
  • Location: Remote-first (U.S.).š Columbus, Ohio is a plus but not required.
  • Reporting: Directly to the founding team, partnering closely with our partner CPAs on domain and review standards.

We are an equal opportunity employer, and we do not discriminate on the basis of race, religion, color, national origin, sex, sexual orientation, age, veteran status, disability, genetic information, or other applicable legally protected characteristic.

Compensation Range: $200K - $300K

Apply for this position

This job is hosted externally. Click below to view the full posting and apply.

Apply on www.indeed.com
Prepare application

Good distractions

Talks and stories from around this role — technically off-topic, practically not.

3:21 min

Navigating the ethical integration of virtual colleagues

Rudi Bauer Rudi Bauer +1 ¡ Cappuccino with HR

3:32 min

Fundamentals and limitations of large language models

Krzystof Czieslak ¡ LIVE

3:46 min

Core terminology and audiences for interpretable artificial intelligence

Karol Przystalski ¡ LIVE

3:33 min

Navigating developer bottlenecks and human accountability

Werner Vogels Werner Vogels +1 ¡ World Congress 2026 Europe

40 sec

Empowering automated workflows with agentic AI models

Mike Mike ¡ World Congress 2025

1:44 min

Balancing artisanal coding skills with automated agent oversight

Jyoti Bansal Jyoti Bansal +1 ¡ World Congress 2026 Europe

Videos

See all

Related articles

See all