AI数据工程师

Staff AI Data Engineer
Location: Remote (U.S.-based) Preference given to CA, TX, and FL
Department: Product
Reports to: Chief Product Officer
Starting Pay Range: $184k - $215k/yrAbout Work Truck Solutions
Work Truck Solutions is the operating system for the commercial vehicle industry. While the retail auto market is saturated with software, the $130B+ commercial truck market was operating in the dark—until we turned on the lights.

We are the only platform that connects the entire ecosystem: OEMs, upfitters, dealerships, and fleet buyers. By digitizing complex inventory data (chassis + upfits) and streamlining the supply chain, we don't just help dealers sell trucks; we ensure American businesses get the mission-critical vehicles they need to work. We are a team of innovators, disruptors, and problem-solvers dedicated to one mission: removing the friction from the commercial vehicle industry. We are profitable, growing, and aggressively scaling our technology to remain the undisputed authority in this space.The Opportunity
We have spent years assembling something no one else in this industry has: a connected view of what's being built, what's sitting on lots, what's moving, and what buyers are looking for and failing to find. Right now, a lot of that value is latent. It lives in our pipelines instead of in our customers' hands.

We are seeking a Staff AI Data Engineer to change that. This role exists to turn our data asset into products dealers, upfitters, and OEMs will pay for and rely on—market intelligence, demand and pricing signal, inventory recommendations, automated enrichment, search and matching that actually finds the right truck.

Your primary job is delivering customer-facing results from our data. That said, you'll need to be able to build the plumbing when it's in your way. The best people for this role don't wait on a ticket queue for a feature pipeline—they architect the ingestion and transformation they need, ship it, and hand it off cleanly. We're looking for someone who is fluent in both directions and clear about which one the moment calls for.

We also want someone genuinely eager about what AI makes possible here. The hardest parts of our data problem—unstructured spec sheets, inconsistent upfit descriptions, entity resolution across feeds that agree on nothing—are exactly the problems where modern AI and LLMs are a step change over what was possible three years ago. You should be excited to reach for those tools, rigorous about proving they worked, and honest when a simpler method wins.How You'll Spend Your Time
Rough shape of the role, so there's no ambiguity about the emphasis:

• ~60% productizing data. Building models, analyses, and data products that reach customers—forecasting, pricing signal, recommendations, matching, enrichment, market intelligence—and iterating on them based on how they actually get used.

• ~25% AI-driven capability. Applying LLMs and ML to extract, structure, and enrich the data that makes those products possible, with the evaluation rigor to know it's working.

• ~15% data engineering. Building the ingestion, transformation, and feature pipelines your work depends on, and setting standards others can follow.

Key Responsibilities
Deliver results from our data

• Own customer-facing data products end to end: define the opportunity, build the model or analysis, ship it, measure whether it actually helped, and iterate.

• Build the intelligence layer of our platform—demand forecasting, pricing and market signal, inventory and configuration recommendations, matching and ranking—on problems where being right has direct commercial consequence for our customers.

• Partner with product and design on how model output surfaces to a dealer or upfitter, what happens when it's wrong, and how much confidence to express.

• Work directly with customers and the commercial team to understand what decisions they're actually trying to make, and let that shape what you build.

• Define and instrument success metrics for everything you ship; be the person who knows whether it worked.

Use AI to unlock the data

• Apply LLMs to the unstructured layer of our business: extraction from spec sheets and vehicle descriptions, classification, taxonomy mapping, enrichment, semantic search and matching.

• Build the evaluation infrastructure that makes AI output trustworthy—golden datasets, offline and online metrics, monitoring for degradation, and guardrails with sensible fallback behavior.

• Bring AI into your own workflow aggressively and critically, and raise the practice of the people around you.

• Make honest calls about where generative approaches beat classical ML or plain deterministic logic, and where they don't.

Build what you need

• Design and build the ingestion, transformation, and feature pipelines your models depend on, rather than waiting for them.

• Contribute to entity resolution and normalization systems that turn inconsistent supplier data into a trustworthy canonical record.

• Establish data quality, lineage, and contract standards for the data your products rest on.

• Partner with data engineering on the platform decisions that outlast any single project, and hand off what you build in a state others can own.

Raise the bar

• Set the standard for analytical and modeling rigor through peer review, and mentor the analysts and engineers around you.

• Write clearly enough that your findings change decisions and your systems can be maintained by someone else.

Qualifications
Data science and productization

• 8+ years applying data science to real problems, with a track record of models and analyses that shipped to users and changed outcomes—not internal reports that circulated and stalled.

• Demonstrated ownership of customer-facing data products, where your model's output was the product and its quality was visible to people paying for it.

• Strong statistical foundation: experimental design, regression, uncertainty quantification, and the judgment to know what your assumptions are and what happens when they break.

• Substantial applied modeling experience across the families that matter here—forecasting, recommendation and ranking, gradient boosting, segmentation, entity resolution and fuzzy matching, anomaly detection.

• Genuine rigor about evaluation: offline metrics that predict online behavior, correct validation for temporal and grouped data, leakage awareness, and calibration—not just accuracy.

• Product instinct. You care whether the customer's decision got better, not whether the model was interesting.

AI fluency and appetite

• Hands-on production experience applying LLMs to data problems: structured extraction, classification and enrichment, embeddings for similarity and clustering, semantic search and matching over messy real-world text.

• Experience building evaluation and guardrail systems for probabilistic output; you don't ship a prompt without a way to know when it degrades.

• Working command of the production tradeoffs—model selection, structured output enforcement, context and token cost, latency, caching, human-in-the-loop review—and able to build a business case that accounts for them.

• Actively curious about the frontier of these tools and eager to apply them, paired with the discipline to verify rather than assume.

• Familiarity with agentic patterns and tool use, with a realistic view of where they're production-ready and where they aren't.

• The judgment to argue against AI when a well-chosen heuristic or a clear dashboard solves the problem more cheaply and more legibly.

Data engineering capability

• Expert SQL and strong Python; you write production code that others maintain comfortably.

• Able to design and build ingestion and ETL/ELT independently—orchestration and transformation tooling, batch and streaming patterns, and sensible data modeling.

• Experience in a modern warehouse or lakehouse environment (BigQuery, Snowflake, Databricks, or equivalent), including awareness of cost and performance.

• Comfortable integrating messy, semi-structured, unreliable third-party sources and building the reconciliation that makes them usable.

• Version control, code review, CI/CD, and reproducible work. Your output is not a folder of untracked notebooks.

Working style

• Exceptional written communication. At this level, the writing is part of the deliverable.

• Demonstrated influence without authority across product, engineering, and commercial teams.

• Comfortable scoping a vague business question into tractable work without being handed the framing.

• Bias toward shipping and learning, with the discipline to follow through past launch.

• Bachelor's degree in a quantitative field, or equivalent depth demonstrated in practice. Advanced degrees welcome but not required.

• Currently resides in one of the following states: CA, TX, FL, MN

Nice to Have

• Experience in automotive, dealership software, logistics, supply chain, fleet, or industrial B2B.

• Pricing, demand forecasting, or inventory optimization background.

• Background with catalog, taxonomy, or configuration data at scale.

• Marketplace experience: liquidity, matching efficiency, supply and demand balance.

• Experience building a company's first customer-facing data product rather than inheriting a mature one.

Why Join Us?

• Remote Flexibility: Work from anywhere in the U.S. while staying connected to a collaborative team.

• Impactful Work: Contribute to products that are reshaping the commercial vehicle industry.

• Growth Opportunities: Be part of a rapidly growing company with ample opportunities for professional development.

• Inclusive Culture: Join a team that values diversity, creativity, and innovation.

Ready to Drive Innovation?

If you're tired of building models that never reach a user—and you want to turn a genuinely unique dataset into products an industry runs on—we'd love to hear from you. Apply now or reach out to us at with your resume.

Originally posted on Himalayas