Senior Applied AI Engineer

Tango; Portland (OR); On-site; USD 160,000 - 190,000; Full time

Benefits offered by this job

Competitive compensation

Comprehensive benefits

Flexible work environment (onsite)

Job summary

Tango in Oregon is hiring a Senior Applied AI Engineer to design, implement, and ship agent and retrieval systems with a focus on evaluation, safety, and collaboration across teams.

The role requires 7+ years in software engineering, including 2+ years building LLM-powered systems in production. Onsite work in Oregon with a salary range of USD 160,000 - 190,000 per year, plus comprehensive benefits.

Qualifications

• 7+ years of professional software engineering experience with 2+ years building LLM-powered systems.
• Strong Python and service stack expertise (FastAPI, Pydantic) with production ownership.
• Experience with agent orchestration frameworks and multi-agent patterns.

Responsibilities

• Design, build, and ship production AI agents on LangGraph with portable cloud deployment.
• Own agent delivery end to end including graph design, tool definitions, prompts, and cost latency budgets.
• Build tooling against MCP and enable agent-to-agent interfaces across systems.
• Create human-in-the-loop review flows with confidence signaling and correction paths.
• Develop retrieval systems: chunking, hybrid retrieval, grounding, and source citation.
• Contribute to agent evaluations using golden datasets and CI-based regression tests.
• Diagnose and fix retrieval, prompt, or model issues to improve reliability and safety.
• Ensure agent safety: prompt-injection resistance, PII handling, and escalation paths.

Skills

Python

FastAPI

Pydantic

Testing

CI/CD

Production ownership

LLM-powered systems

Agent orchestration

Tools

LangGraph

LangChain

OpenAI

Claude Agents SDKs

MCP tool servers

LangSmith

Langfuse

Arize

Job description

Tango is building its first AI-powered product, and the Senior Applied AI Engineer role is central to turning generative AI and machine learning into capabilities that can be relied on in production. You will design, implement, and ship agent and retrieval systems, with an emphasis on evaluation, safety, and collaboration across teams.

Based in Oregon for onsite work, this position offers a salary range of USD 160,000 - 190,000 per year and requires 7+ years of professional software engineering experience, including hands‑on work on LLM‑powered systems used by real customers.

Responsibilities

• Design, build, and ship production AI agents on LangGraph, keeping the agent layer portable across cloud platforms.
• Own agent delivery end to end, including graph design, tool definitions, prompt and context engineering, durable execution, failure and retry behavior, and cost and latency budgets.
• Build agent tooling against internal systems over MCP, using direct API calls where they fit and enabling agent-to-agent interfaces as systems compose.
• Create human-in-the-loop review flows with interrupt points, confidence surfacing, and correction paths that let customers review and override agent output.
• Build and tune retrieval, including chunking, hybrid retrieval, grounding, and citation back to source page and paragraph.
• Contribute to agent evaluations using golden datasets, LLM-as-judge and deterministic scorers, and regression suites that run in CI using the platform’s shared evaluation harness.
• Diagnose quality issues to root cause, such as retrieval misses, prompt defects, tool errors, model regressions, or flawed ground truth, then fix the correct layer.
• Own agent-level safety behavior, including prompt-injection resistance, PII handling, and refusal and escalation paths, applying the platform guardrail service maintained by Platform Engineering.
• Partner with Product to translate accuracy thresholds, confidence disclosure, and human-in-the-loop triggers into shipped behavior.
• Coordinate with Platform Engineering on deployment and with Data Platform on the curated datasets agents read.
• Feed curated agent session and usage analytics into the warehouse so agent performance can be measured alongside product analytics.
• Transition reference agents to domain teams for long-term operation, and help define a shared agent quality standard.

Requirements

• 7+ years of professional software engineering experience, including 2+ years building LLM-powered systems that reached production and real users.
• Strong expertise in Python and its service stack (including FastAPI and Pydantic or equivalents), plus engineering practices such as testing, code review, CI/CD, and production ownership.
• Production experience with an agent orchestration framework; LangGraph is strongly preferred, while LangChain, OpenAI or Claude Agents SDKs, or equivalent frameworks are considered.
• Hands‑on depth with at least one frontier model API.
• Hands‑on experience with LLM evaluation, including golden datasets, LLM-as-judge and deterministic scorers, regression testing, and using evaluation results to gate releases; evaluation is central to the role.
• Experience with MCP tool servers or comparable tool and function‑calling protocols, along with multi‑agent patterns.
• Production RAG and retrieval experience, including chunking strategy, hybrid retrieval, grounding, citation, and diagnosing retrieval failures.
• Experience with LLM observability and tracing (including LangSmith, Langfuse, Arize, or equivalents), and with prompt and version management.
• Sound judgment about failure modes including hallucination, prompt injection, and silent degradation, with the ability to distinguish acceptable from unacceptable failures.

Technologies

• Python, FastAPI, Pydantic
• LangGraph, LangChain, OpenAI, Claude Agents SDKs
• MCP, MCP tool servers
• LangSmith, Langfuse, Arize
• CI/CD
• RAG, LLM-as-judge
• Vector and hybrid retrieval stores: pgvector, Pinecone, Weaviate, Qdrant
• Celery/Redis

Benefits

• Competitive Compensation
• Comprehensive Benefits (health, dental, and vision insurance)
• 401(k) plan with company match
• Generous paid time off
• Flexible work environment (remote, hybrid, or in-office)
• Inclusive & Collaborative Culture

Preferred

• Graph-backed agent memory or knowledge graphs (Neo4j or similar).
• Production vector and hybrid retrieval stores (pgvector, Pinecone, Weaviate, Qdrant).
• Async task orchestration for long-running document pipelines (Celery/Redis or equivalent).
• Document intelligence and information extraction at scale, including OCR, layout‑aware parsing, and structured extraction from long documents.