Staff AI Engineering - Enterprise Architecture
American Express · New York, NY · 3 wk ago
HybridEngineeringFull-time
We need to output HTML fragment with sections: About the role, Responsibilities, Requirements (or Qualifications), Skills, Benefits, Pay, Schedule. Only include sections that have real specific content. The posting includes: About the role (the "The Role" paragraph). Responsibilities (list). Qualifications (list). Skills (list of languages, APIs, etc). Benefits (list). Pay: there is a paragraph about salary range but not actual numbers; says "The below represents the expected salary range for this job requisition." but no actual numbers given. It says "Ultimately, in determining your pay, we’ll consider your location, experience, and other job-related factors." No specific numbers. So we should NOT include Pay section because no specific content. Schedule: mentions flexible working model, hybrid, onsite or virtual arrangements depending on role and business need. That's schedule info. So we can include Schedule section with that content.
We must not repeat job title as heading. Use for section headings. Use
for section headings. Use - for any list of duties/requirements/benefits. Use
for prose.
We need to preserve all substantive information, not invent or summarize away detail. Remove boilerplate etc.
We need to include sections only when posting has real specific content.
Let's extract content:
About the role: The paragraph after "The Role As a Staff Software Engineer – Agentic AI, you will operate as a senior individual contributor..." includes details. We'll put that in
.
Responsibilities: The bullet list under "Responsibilities What You’ll Do". It includes many items, some are paragraphs, some are bullet-like lines. We'll convert each line to
- . Need to preserve all details. The list includes:
- Core engineering stack
Drive technical direction for agentic AI initiatives, influencing architecture patterns, autonomy boundaries, and system design.
Design, build, and operate production-grade agentic AI systems used across multiple products.
Own and evolve shared agentic AI capabilities, including:
Agent frameworks and orchestration layers
Planning, tool use, and memory strategies
Retrieval and grounding (RAG) pipelines
LLM infrastructure, inference, and model gateways
Evaluation, observability, and safety tooling for autonomous systems
Lead technical design reviews and help teams navigate tradeoffs involving autonomy, safety, reliability, scalability, and cost.
Partner across teams to deliver complex, cross-cutting agentic AI initiatives from concept to production.
Evaluate emerging models, techniques, and agentic patterns and translate them into practical, enterprise-ready improvements.
Mentor senior engineers and raise the technical bar for agentic AI development through example and influence.
Languages: Python, Go, TypeScript
APIs and RPC: REST, gRPC (and in some areas tRPC)
Cloud and infrastructure: AWS and/or GCP, Kubernetes
Distributed systems: event-driven architectures, including Kafka
Agentic AI and ML Integration of commercial and open-source LLMs into agentic workflows
Agent and orchestration frameworks such as LangChain, LlamaIndex, Semantic Kernel, or CrewAI, with strong judgment about when to use frameworks versus building lighter-weight primitives
Model-level work using PyTorch and the Hugging Face ecosystem (embeddings, fine-tuning, inference tooling), with some exposure to TensorFlow
Strong schema, validation, and state management practices using tools such as Pydantic (Python) and Zod (TypeScript)
Across all systems, we emphasize evaluation, observability, safety, and reliability, reflecting the responsibility of deploying autonomous AI in a regulated, customer-facing environment.
We need to decide how to structure list items. Could treat each bullet as
- . Might be long but okay.
Qualifications: The bullet list under "Qualifications". It includes repeated line: "+10 years of software engineering expereince, including deep, hands-on experience delivering agentic or LLM-powered systems to production" appears twice (maybe a mistake). We need to preserve all substantive information; we can include both but maybe deduplicate? The instruction: "Preserve ALL substantive information; do NOT invent or summarize away detail." If there's duplicate, we should keep it as is? Might be okay to keep both but could be considered summarizing away detail if we remove duplicate. Safer to keep both entries.
List:
- +10 years of software engineering expereince, including deep, hands-on experience delivering agentic or LLM-powered systems to production
- Proven ownership of complex, cross-cutting agentic systems spanning multiple teams or products.
- +10 years of software engineering expereince, including deep, hands-on experience delivering agentic or LLM-powered systems to production (duplicate)
- Strong engineering fundamentals across backend systems, APIs, data pipelines, and cloud infrastructure.
- Deep experience across the agentic AI stack, including planning, tool use, memory, and evaluation.
- Fluency with AI-assisted and agentic development workflows.
- Comfort operating in ambiguous problem spaces and translating them into shipped, reliable autonomous systems.
- Ability to influence technical direction and align teams without formal authority.
- Experience in worflow engines, async processing, queues, and streaming systems
Preferred Qualifications: list:
- Experience building agentic systems in fintech or other regulated industries.
- Experience as a founding engineer or early technical leader in AI-driven products.
- Demonstrated success delivering technically complex autonomous systems that customers actively rely on.
- Meaningful contributions to open-source AI or agentic frameworks.
- Familiarity with fine-tuning, model optimization and inference pipelines is a plus.
We could combine Qualifications and Preferred Qualifications into a single "Requirements" section? The instruction says use headings like Requirements, Qualifications, Skills, Benefits, Pay, Schedule. We have both Qualifications and Preferred Qualifications; we could put them under "Qualifications" heading, with sub-items? But we need to follow instruction: Use
for section headings (e.g. About the role, Responsibilities, Requirements, Qualifications, Skills, Benefits, Pay, Schedule). So we can have both Qualifications and maybe Requirements? The posting has "Qualifications" and "Preferred Qualifications". We could treat "Qualifications" as heading and include both required and preferred items. Or we could have "Requirements" for required and "Preferred Qualifications" as separate heading? The instruction list includes Requirements and Qualifications as separate possible headings. We could use both: Qualifications
for required, and Preferred Qualifications
for preferred. That seems okay.
Skills: The posting includes languages, APIs, etc. That's under Responsibilities but also could be a Skills section. We'll extract those items into a Skills section.
Benefits: The "About The Team" paragraph lists benefits. We'll put that as a list.
Schedule: The flexible working model description.
Now we need to remove boilerplate: cookie/navigation/legal/EEO/equal-opportunity/scam-warning boilerplate, duplicated lines, and application instructions. Also remove "About Us", "About The Team" maybe we keep benefits but not the fluff. The "About Us" paragraph is mostly fluff; we can keep benefits but not the generic culture description. The instruction: "Prefer a few substantive sections over many thin ones." So we might combine some.
We need to output ONLY the HTML fragment.
Let's craft:
About the role
As a Staff Software Engineer – Agentic AI, you will operate as a senior individual contributor within Amex Technology, helping define how agentic AI systems are designed, built, and operated across the company. You will work at the intersection of agentic system architecture and hands-on execution: designing core frameworks, writing production code, reviewing critical designs, and guiding teams through complex technical decisions. This role does not include people management; your impact comes from technical leadership, sound judgment, and the ability to turn ambiguity into reliable, scalable agentic systems. You will work closely with Product and UX partners to shape a shared vision for agentic AI experiences. While product teams own prioritization, Staff engineers are expected to actively influence what gets built and how agentic systems come together across teams. We are primarily looking for senior individual contributors; however, as the team grows, you will take on increasing responsibility for technical leadership and people development within the agentic AI space. Mentoring, developing and managing direct reports may be in scope for some of our opportunities.
Responsibilities
- Drive technical direction for agentic AI initiatives, influencing architecture patterns, autonomy boundaries, and system design.
- Design, build, and operate production-grade agentic AI systems used across multiple products.
- Own and evolve shared agentic AI capabilities, including:
- Agent frameworks and orchestration layers
- Planning, tool use, and memory strategies
- Retrieval and grounding (RAG) pipelines
- LLM infrastructure, inference, and model gateways
- Evaluation, observability, and safety tooling for autonomous systems
- Lead technical design reviews and help teams navigate tradeoffs involving autonomy, safety, reliability, scalability, and cost.
- Partner across teams to deliver complex, cross-cutting agentic AI initiatives from concept to production.
- Evaluate emerging models, techniques, and agentic patterns and translate them into practical, enterprise-ready improvements.
- Mentor senior engineers and raise the technical bar for agentic AI development through example and influence.
- Languages: Python, Go, TypeScript
- APIs and RPC: REST, gRPC (and in some areas tRPC)
- Cloud and infrastructure: AWS and/or GCP, Kubernetes
- Distributed systems: event-driven architectures, including Kafka
- Agentic AI and ML: Integration of commercial and open-source LLMs into agentic workflows
- Agent and orchestration frameworks such as LangChain, LlamaIndex, Semantic Kernel, or CrewAI, with strong judgment about when to use frameworks versus building lighter-weight primitives
- Model-level work using PyTorch and the Hugging Face ecosystem (embeddings, fine-tuning, inference tooling), with some exposure to TensorFlow
- Strong schema, validation, and state management practices using tools such as Pydantic (Python) and Zod (TypeScript)
- Across all systems, we emphasize evaluation, observability, safety, and reliability, reflecting the responsibility of deploying autonomous AI in a regulated, customer-facing environment.
But note we already included languages etc inside responsibilities; maybe better to move those to Skills section. However the instruction: Use - for any list of duties/requirements/benefits. The languages etc are more like skills, not duties. Could still be in responsibilities but we can separate.
Let's create Skills section for those technical items.
Thus responsibilities list will be the duties (excluding the technical stack list). Then Skills section will list languages, APIs, cloud, etc.
Let's extract duties:
- Drive technical direction for agentic AI initiatives, influencing architecture patterns, autonomy boundaries, and system design.
- Design, build, and operate production-grade agentic AI systems used across multiple products.
- Own and evolve shared agentic AI capabilities, including:
- Agent frameworks and orchestration layers
- Planning, tool use, and memory strategies
- Retrieval and grounding (RAG) pipelines
- LLM infrastructure, inference, and model gateways
- Evaluation, observability, and safety tooling for autonomous systems
- Lead technical design reviews and help teams navigate tradeoffs involving autonomy, safety, reliability, scalability, and cost.
- Partner across teams to deliver complex, cross-cutting agentic AI initiatives from concept to production.
- Evaluate emerging models, techniques, and agentic patterns and translate them into practical, enterprise-ready improvements.
- Mentor senior engineers and raise the technical bar for agentic AI development through example and influence.
We can keep the nested list for owned capabilities as sub-items.
Now Skills section:
- Languages: Python,
for prose. We need to preserve all substantive information, not invent or summarize away detail. Remove boilerplate etc. We need to include sections only when posting has real specific content. Let's extract content: About the role: The paragraph after "The Role As a Staff Software Engineer – Agentic AI, you will operate as a senior individual contributor..." includes details. We'll put that in
. Responsibilities: The bullet list under "Responsibilities What You’ll Do". It includes many items, some are paragraphs, some are bullet-like lines. We'll convert each line to
for section headings (e.g. About the role, Responsibilities, Requirements, Qualifications, Skills, Benefits, Pay, Schedule). So we can have both Qualifications and maybe Requirements? The posting has "Qualifications" and "Preferred Qualifications". We could treat "Qualifications" as heading and include both required and preferred items. Or we could have "Requirements" for required and "Preferred Qualifications" as separate heading? The instruction list includes Requirements and Qualifications as separate possible headings. We could use both: Qualifications
for required, and Preferred Qualifications
for preferred. That seems okay.
Skills: The posting includes languages, APIs, etc. That's under Responsibilities but also could be a Skills section. We'll extract those items into a Skills section.
Benefits: The "About The Team" paragraph lists benefits. We'll put that as a list.
Schedule: The flexible working model description.
Now we need to remove boilerplate: cookie/navigation/legal/EEO/equal-opportunity/scam-warning boilerplate, duplicated lines, and application instructions. Also remove "About Us", "About The Team" maybe we keep benefits but not the fluff. The "About Us" paragraph is mostly fluff; we can keep benefits but not the generic culture description. The instruction: "Prefer a few substantive sections over many thin ones." So we might combine some.
We need to output ONLY the HTML fragment.
Let's craft:
About the role
As a Staff Software Engineer – Agentic AI, you will operate as a senior individual contributor within Amex Technology, helping define how agentic AI systems are designed, built, and operated across the company. You will work at the intersection of agentic system architecture and hands-on execution: designing core frameworks, writing production code, reviewing critical designs, and guiding teams through complex technical decisions. This role does not include people management; your impact comes from technical leadership, sound judgment, and the ability to turn ambiguity into reliable, scalable agentic systems. You will work closely with Product and UX partners to shape a shared vision for agentic AI experiences. While product teams own prioritization, Staff engineers are expected to actively influence what gets built and how agentic systems come together across teams. We are primarily looking for senior individual contributors; however, as the team grows, you will take on increasing responsibility for technical leadership and people development within the agentic AI space. Mentoring, developing and managing direct reports may be in scope for some of our opportunities.
Responsibilities
- Drive technical direction for agentic AI initiatives, influencing architecture patterns, autonomy boundaries, and system design.
- Design, build, and operate production-grade agentic AI systems used across multiple products.
- Own and evolve shared agentic AI capabilities, including:
- Agent frameworks and orchestration layers
- Planning, tool use, and memory strategies
- Retrieval and grounding (RAG) pipelines
- LLM infrastructure, inference, and model gateways
- Evaluation, observability, and safety tooling for autonomous systems
- Lead technical design reviews and help teams navigate tradeoffs involving autonomy, safety, reliability, scalability, and cost.
- Partner across teams to deliver complex, cross-cutting agentic AI initiatives from concept to production.
- Evaluate emerging models, techniques, and agentic patterns and translate them into practical, enterprise-ready improvements.
- Mentor senior engineers and raise the technical bar for agentic AI development through example and influence.
- Languages: Python, Go, TypeScript
- APIs and RPC: REST, gRPC (and in some areas tRPC)
- Cloud and infrastructure: AWS and/or GCP, Kubernetes
- Distributed systems: event-driven architectures, including Kafka
- Agentic AI and ML: Integration of commercial and open-source LLMs into agentic workflows
- Agent and orchestration frameworks such as LangChain, LlamaIndex, Semantic Kernel, or CrewAI, with strong judgment about when to use frameworks versus building lighter-weight primitives
- Model-level work using PyTorch and the Hugging Face ecosystem (embeddings, fine-tuning, inference tooling), with some exposure to TensorFlow
- Strong schema, validation, and state management practices using tools such as Pydantic (Python) and Zod (TypeScript)
- Across all systems, we emphasize evaluation, observability, safety, and reliability, reflecting the responsibility of deploying autonomous AI in a regulated, customer-facing environment.
- for any list of duties/requirements/benefits. The languages etc are more like skills, not duties. Could still be in responsibilities but we can separate. Let's create Skills section for those technical items. Thus responsibilities list will be the duties (excluding the technical stack list). Then Skills section will list languages, APIs, cloud, etc. Let's extract duties: - Drive technical direction for agentic AI initiatives, influencing architecture patterns, autonomy boundaries, and system design. - Design, build, and operate production-grade agentic AI systems used across multiple products. - Own and evolve shared agentic AI capabilities, including: - Agent frameworks and orchestration layers - Planning, tool use, and memory strategies - Retrieval and grounding (RAG) pipelines - LLM infrastructure, inference, and model gateways - Evaluation, observability, and safety tooling for autonomous systems - Lead technical design reviews and help teams navigate tradeoffs involving autonomy, safety, reliability, scalability, and cost. - Partner across teams to deliver complex, cross-cutting agentic AI initiatives from concept to production. - Evaluate emerging models, techniques, and agentic patterns and translate them into practical, enterprise-ready improvements. - Mentor senior engineers and raise the technical bar for agentic AI development through example and influence. We can keep the nested list for owned capabilities as sub-items. Now Skills section: - Languages: Python,