AI Engineer

Better Direct AI; Tempe (AZ); On-site; USD 120,000 - 160,000; Full time

Benefits offered by this job

401(k)

Dental insurance

Health insurance

Life insurance

Paid time off

Professional development assistance

Retirement plan

Vision insurance

Job summary

Better Direct AI in Tempe, AZ, is seeking an AI Engineer to own AI capabilities from concept to production, including models, data pipelines, APIs, and UX. You’ll focus on document transformation and workflow automation to move quotes and product data through production systems with quality controls.

You will design, deploy, and monitor end-to-end AI components, build Python services with FastAPI, and integrate AI outputs into CRM workflows with audit trails and confidence scores.

Qualifications

• 1+ year of experience building applications powered by LLMs or other machine-learning models used in real workflows.
• Strong Python skills, including FastAPI and asynchronous programming.
• Working knowledge of TypeScript or JavaScript, with REST APIs or React experience.
• Strong SQL and PostgreSQL fundamentals: schema design, migrations, indexing, and queries.
• Hands-on experience in two areas: RAG/embeddings/semantic search, OCR, model serving, or similar.

Responsibilities

• Design document-processing pipelines converting RFQ PDFs, scanned pages, spreadsheets, and emails into structured outputs.
• Run, optimize, and monitor self-hosted LLMs, vision-language models, OCR systems, and embeddings on on-prem GPUs and Kubernetes.
• Build search and RAG systems for contracts, catalogs, and quotes using PostgreSQL, pgvector, BM25, and vector retrieval.
• Develop tool-using agents that research products across distributor catalogs and prepare quotes for review.
• Train and evaluate classifiers, NER models, and sequence-labeling models when LLMs are slow or expensive.
• Create labeled evaluation datasets, regression tests, and performance benchmarks for accuracy and latency.
• Build Python services using FastAPI, Celery, and Redis for extraction, crawling, polling, and background workloads.
• Ship AI capabilities into the CRM using Express and Prisma APIs, PostgreSQL migrations, and React interfaces.
• Integrate AI outputs into production workflows with confidence scores, human review, and audit trails.
• Deploy and maintain services with Docker and Git, using runbooks and post-release monitoring.

Skills

LLM applications

Python

TypeScript/JavaScript

SQL PostgreSQL

Linux/Git/Docker

Communication

Tools

FastAPI

Kubernetes

Docker

Redis

Celery

Express

Prisma

Selenium

Playwright

Job description

Better Direct AI builds AI-powered tools that turn messy business inputs into reliable workflows for sales and engineering teams. As an AI Engineer in Tempe, AZ, you will own AI capabilities from concept to production, covering models, data pipelines, APIs, and user-facing experiences. You will focus on document transformation and workflow automation, helping move quotes and product data through production systems with quality controls built in.

What you’ll do

• Design document-processing pipelines that convert RFQ PDFs, scanned pages, spreadsheets, and emails into structured outputs such as part numbers, quantities, manufacturers, and shipping addresses.
• Run, optimize, and monitor self-hosted LLMs, vision-language models, OCR systems, embedding models, and smaller task-specific models on on-premises NVIDIA GPUs and a Kubernetes-based AI cluster.
• Build search and RAG systems for contracts, product catalogs, and historical quotes using PostgreSQL, pgvector, BM25, vector retrieval, reranking, and deduplication.
• Develop tool-using agents that research products across distributor catalogs (including TD SYNNEX, Ingram Micro, and D&H) and prepare quotes for human review.
• Train and evaluate classifiers, NER models, sequence-labeling models, and other efficient alternatives when large language models are too slow or expensive.
• Create labeled evaluation datasets, regression tests, and performance benchmarks that track accuracy, recall, latency, reliability, and cost.
• Build and maintain Python services using FastAPI, asynchronous processing, Celery, and Redis for extraction, crawling, polling, and other background workloads.
• Ship AI capabilities into the CRM using Express and Prisma APIs, PostgreSQL schema changes and migrations, and React interfaces.
• Integrate AI outputs into production workflows with confidence scores, human review, feedback capture, permissions, and audit trails.
• Deploy and maintain services with Docker and Git, using documented runbooks, and monitor production jobs to diagnose failures and support post-release improvements.
• Write pull requests, design notes, test plans, and handoff documentation so other engineers can understand and maintain your work.

What you bring

• U.S. citizenship.
• At least 1 year of experience building applications powered by LLMs or other machine-learning models used in real workflows.
• Strong Python skills, including FastAPI (or a similar framework), asynchronous programming, packaging, and automated testing.
• Working knowledge of TypeScript or JavaScript, with experience building REST APIs or React applications, or the ability to learn quickly.
• Strong SQL and PostgreSQL fundamentals: schema design, migrations, indexing, and query optimization.
• Hands-on experience in at least two areas: RAG/embeddings/semantic search, document extraction or OCR, agent tools/prompting/workflow orchestration, model fine-tuning or training, model serving with vLLM or SGLang (or similar), creating labeled test sets and objective metric evaluation, comfort with Linux/Git/Docker.
• Clear written and verbal communication, including the ability to explain technical trade-offs to non-technical stakeholders.

Benefits

• 401(k)
• Dental insurance
• Health insurance
• Life insurance
• Paid time off
• Professional development assistance
• Retirement plan
• Vision insurance

AI-assisted engineering

AI-assisted development is part of how the team works. Engineers may use tools such as OpenAI Codex to research systems, explore solutions, write code, generate tests, debug problems, and produce documentation. Tools support the work, but the engineer remains responsible for the final result.

• Translate unclear business requests into defined technical tasks with acceptance criteria.
• Provide AI tools with the right context, constraints, and technical requirements.
• Read, understand, explain, and maintain all code you submit.
• Identify incomplete, unsafe, inaccurate, or unnecessarily complex generated solutions.
• Verify assumptions against the existing codebase, database, architecture, and application behavior.
• Evaluate trade-offs and choose practical, maintainable solutions.
• Test permissions, security, data integrity, performance, failure modes, and edge cases.
• Use approved tools responsibly with company and customer data.
• Troubleshoot and support your work after release.

Preferred qualifications

• Experience running open-weight models such as Qwen, Llama, or Mistral on self-managed GPUs.
• Knowledge of quantization, multi-GPU inference, model optimization, and NVIDIA driver or CUDA troubleshooting.
• Kubernetes experience, particularly with KServe or another model-serving platform.
• Experience with vision-language models or OCR frameworks such as Docling.
• Background in traditional NLP techniques such as text classification, named-entity recognition, active learning, weak supervision, or conditional random fields.
• Experience using Celery and Redis, Selenium or Playwright, or S3-compatible storage such as MinIO.
• Familiarity with IT hardware and software products, manufacturer part numbers, and distributor catalogs.
• Contributions to open-source machine-learning or developer-tooling projects.

Technologies you may work with

Python, FastAPI, TypeScript, JavaScript, REST, React, SQL, PostgreSQL, pgvector, BM25, PostgreSQL schema changes and migrations, Express, Prisma, Docker, Git, Linux, Kubernetes, NVIDIA GPUs, Celery, Redis, vLLM, SGLang, OpenAI Codex, OCR, embeddings, RAG, vector retrieval, reranking, deduplication, TD SYNNEX, Ingram Micro, D&H, KServe, Docling, Selenium, Playwright, MinIO, CUDA.

What we value

• Systems thinking: how models, services, databases, interfaces, and integrations work together.
• Data correctness: preventing duplicate, missing, inconsistent, or unauthorized changes.
• Practical problem-solving: identifying root causes and comparing solutions based on evidence.
• Production ownership: validating releases, monitoring results, and responding when problems occur.
• Scalability and maintainability: anticipating designs that may become slow, fragile, or hard to extend.
• Business understanding: learning why a workflow exists, not just how to implement it.
• Good judgment: knowing when to proceed, investigate further, simplify, or ask for help.
• Continuous learning: developing new skills quickly and taking on greater technical ownership over time.