Posted 2 months ago
For companies:
For professionals:
For everyone:
For developers:
Responsibilities
Rails Platform Engineering:
- Build and ship product features across a Ruby on Rails backend hosted on Heroku and two React applications hosted on Vercel.
- Work within a mature, layered monolith that supports a real-world physical operation, including state machines, warehouse workflows, feature flags, device records, and third-party integrations.
- Improve the architecture incrementally while supporting a roadmap of new products and operational capabilities.
- Balance AI development with the reliability, maintainability, and performance expectations of the core Rails platform.
- Contribute across the full development lifecycle, including implementation, testing, deployment, production support, and observability.
AI Platform, Evaluation, and Guardrails:
- Build the shared AI infrastructure used across product features and internal engineering workflows.
- Create evaluation harnesses that measure model performance against real company data and known outcomes before changes reach production.
- Establish guardrails for model inputs, outputs, tool use, confidence thresholds, and human escalation.
- Design AI systems that fail safely when data is incomplete, ambiguous, or inconsistent.
- Track model behavior in production and turn human corrections, overrides, and exceptions into measurable feedback.
- Define shared grading and evaluation standards so different AI-enabled features are assessed consistently.
- Help the company maintain control over its AI systems as they move from prototypes into operational workflows.
Device Data, Serial Numbers, and Grading:
- Build workflows that use serial numbers and other device identifiers to retrieve and enrich device information.
- Connect internal device records with trusted external or proprietary data sources while preserving traceability and data quality.
- Normalize inconsistent device information collected from customers, warehouses, integrations, and historical records.
- Develop systems that classify or grade devices using the company’s A–D rating framework.
- Train, fine-tune, or otherwise adapt models using the company’s historical dataset when that provides a measurable advantage over prompting or retrieval alone.
- Design confidence scoring and review workflows for cases where an identifier cannot be reliably resolved.
- Protect inventory workflows from small but consequential errors, such as confusing the letter “O” with the number “0” in a serial number.
AI-Assisted Customer Support:
- Build triage and classification systems that automatically route incoming customer requests.
- Implement semantic deduplication so recurring questions can be answered from established precedent rather than repeatedly reaching a person.
- Ground suggested responses in answers the company has previously reviewed and provided.
- Automatically create Linear issues when a support request is determined to be a product defect, including the relevant customer and technical context.
- Design clear escalation paths to human operators and measure how frequently and why escalation occurs.
- Ensure AI-generated responses are auditable, appropriately sourced, and easy for a person to review.
Retrieval and Reasoning from Historical Data:
- Move AI systems from relying only on static rubrics toward reasoning from relevant historical precedent.
- Build retrieval workflows over years of data involving devices, grades, processing decisions, and resale outcomes.
- Determine when relational queries, search, embeddings, or other retrieval techniques are appropriate.
- Make pragmatic infrastructure decisions based on the problem rather than introducing technologies by default.
- Evaluate whether retrieved examples meaningfully improve model accuracy and decision quality.
Agents That Act:
- Build controlled agents capable of performing actions within the company’s domain, such as looking up a device, checking a shipment, updating a record, or drafting a customer update.
- Define which actions can run automatically and which require explicit human approval.
- Implement authorization, validation, audit logging, rollback strategies, and operational safeguards.
- Prevent agents from taking actions outside their intended scope.
- Extend agent-based workflows into engineering tasks where appropriate, including code generation, test creation, change suggestions, and code review.
AI-Enabled Development Workflows:
- Identify practical opportunities to use AI throughout software development, QA, and DevOps.
- Improve how engineers use AI for implementation, debugging, test generation, documentation, incident investigation, and code review.
- Develop shared workflows and reusable context so engineers do not repeatedly solve the same prompting or tooling problems in isolation.
- Evaluate AI-assisted engineering tools based on measurable improvements in quality, speed, or reliability.
- Ensure AI-generated code and infrastructure changes follow the same review, testing, and security standards as human-written work.
- Help engineers build confidence in AI tools without encouraging uncritical reliance on them.
AI Adoption and Team Enablement:
- Help the organization adopt AI through practical delivery rather than presentations or isolated experiments.
- Work consultatively with engineers, product leaders, and operational teams to identify appropriate AI use cases.
- Address understandable hesitation by making model behavior visible, measurable, and controllable.
- Explain the limitations and tradeoffs of proposed AI systems to technical and nontechnical stakeholders.
- Make AI fluency a shared team capability rather than knowledge concentrated in one person.
- Establish patterns, documentation, and examples that other engineers can follow.
- Raise the engineering standard by demonstrating disciplined, production-oriented AI development.
Required Experience:
- 5+ years of production Ruby on Rails experience, including significant experience working within a large, layered monolithic application.
- Deep Rails knowledge and a track record of being the engineer others rely on for difficult framework, architecture, and production issues.
- Hands-on experience shipping LLM or machine-learning features to production.
- Experience with several applied AI patterns, such as retrieval, structured outputs, multimodal models, classification, tool use, agents, or model adaptation.
- Experience building evaluation harnesses for AI systems, not only implementing AI-powered features.
- Experience designing guardrails, confidence thresholds, human-review workflows, and safe failure modes.
- Experience working with inconsistent or incomplete datasets, including normalization, entity resolution, or classification.
- Sound judgment about when prompting, retrieval, fine-tuning, traditional software, or a non-AI solution is most appropriate.
- Demonstrated ability to identify situations where AI should not be used.
- Comfort working across the stack, including enough React experience to test, troubleshoot, and complete work in the company’s frontend applications.
- Familiarity with production deployment, observability, testing, and operational support.
- At least 75% overlap with Eastern Time.
Valuable Additional Experience:
- Experience training, fine-tuning, or adapting models using proprietary company datasets.
- Experience building systems involving serial numbers, product catalogs, hardware identifiers, asset data, or entity matching.
- Experience with AI-assisted QA, DevOps, code review, or software-development workflows.
- Experience integrating LLMs with internal tools, APIs, databases, or transactional systems.
- Experience operating AI systems where errors can affect inventory, logistics, financial outcomes, or physical operations.
- Experience introducing AI capabilities into an organization that is still developing its adoption strategy.
Professional Attributes:
- You are an engineer first and prefer shipping working software over producing theoretical AI strategies.
- You default to measurement.
- You understand that a convincing demonstration is not the same as a reliable production system.
- You know when to say no to AI and can explain why a deterministic or human-driven solution would be safer or more effective.
- You are pragmatic about infrastructure and prefer using the existing stack before introducing new services or databases.
- You are comfortable moving between Rails product development, AI experimentation, data analysis, and production operations.
- You think carefully about data quality, edge cases, ambiguity, and failure modes.
- You are happiest with your hands on the keyboard.
- You are comfortable in a Rails console, willing to read gem source when documentation falls short, and experienced enough with production systems to know where failures tend to hide.
- You can work consultatively with stakeholders who may be uncertain or cautious about adopting AI.
- You are comfortable being the most knowledgeable person on a topic without needing to be the loudest person in the room.
- You raise the capabilities of the people around you through examples, shared tools, documentation, and thoughtful technical leadership.
This job post is closed.