Principal Architect

Sutherland

3d

About The Position

We are seeking a Principal Architect to design and build agentic harnesses from scratch, focusing on capabilities like tool use, multi-step chaining, reasoning, streaming, skills, multimodal integration, RAG, sandboxing, and state management. The role involves designing, building, and operating production APIs and services, including RESTful APIs, streaming APIs, asynchronous workflows, service boundaries, versioning, authentication/authorization, error handling, and backward compatibility. You will demonstrate ownership of production systems, contributing to process improvements, roadmap development, defect resolution, and ensuring uptime and availability. Strong backend engineering fundamentals are essential, with comfort across service design, APIs, batch workflows, orchestration, observability, and production support. Familiarity with evaluation techniques for AI systems, including creating and maintaining evaluation datasets, running offline regression evaluations, monitoring online production performance, rubric-based scoring, self-verification, human-in-the-loop review, AI-as-judge methods, and quality/reliability analysis at scale, is crucial. A clear understanding of foundational machine learning and statistical concepts, including sampling, statistical significance, overfitting/underfitting, precision and recall, and quality tradeoff analysis, is required. The role also involves building and orchestrating large-scale batch workflows using foundation models (LLMs, VLMs), pretrained open-source models, and deep learning models. You will design systems that safely and reliably automate enterprise workflows using non-deterministic AI components, with strong judgment around reliability, failure handling, observability, human review, and operational safety. A deep understanding of making systematic tradeoffs between quality, reliability, latency, cost, explainability, and user experience in complex AI systems is expected. The ability to design pragmatic architectures that balance innovation with production readiness and comfort working in environments with non-deterministic model behavior, accounting for uncertainty, evaluation, monitoring, and graceful failure, are key. Experience with both lexical and embedding-based search methods, reasoning about relevance, ranking, latency, recall, precision, indexing strategy, and retrieval performance is necessary. Experience working with foundation models and open-source LLMs beyond simple API calls, and familiarity with lower-level model behaviors and controls (temperature, top-p sampling, logprobs, confidence scoring, prompting strategies, model selection tradeoffs) are important. While the role is primarily backend and AI systems focused, comfort and willingness to build front-end experiences using JavaScript/TypeScript and frameworks like React, and to work across the stack to own the end-to-end product experience, is essential. Familiarity with tools such as FFmpeg, OpenCV, or similar media-processing libraries is a plus.

Requirements

• 7+ years of a programming languages such as Java, Golang, Python, JavaScript
• 2+ years of experience architecting, designing, and optimizing search and information retrieval systems at scale.
• 6 months to 1 year of hands-on experience building agentic products or solutions, including tool-using agents, conversational agents, long-running agents, reasoning/planning agents, or similar systems.
• 1+ year of experience evaluating AI systems for quality, reliability, and safety at scale.
• 5+ years of experience shipping production software in an enterprise or consumer environment, not just prototypes or proofs of concept.
• Expert experience, understanding and knowledge of digital and broadcast production operations and workflows
• Experience working with video, rich visual media, or multimodal AI systems.
• Strong knowledge of industry trends and best practices
• Strong experience with the C4 model and other traditional design artifacts
• Experience working in an Agile environment
• BS in CS, EE or equivalent experience required

Nice To Haves

• Experience with multiple modern agent frameworks is preferred. Specific frameworks are less important than demonstrated fluency, but exposure to tools such as OpenAI Agents SDK, LangGraph, Google ADK, SmolAgents, or comparable frameworks is valuable.
• OpenSearch or Elasticsearch experience is preferred, but comparable experience with other search and retrieval systems is sufficient.
• Experience with training, fine-tuning, or adapting models for domain-specific use cases using techniques such as LoRA, PEFT, or related approaches.
• Exposure to stateful software design patterns and streaming protocols such as WebSockets
• Familiarity with client-side web technologies (React, Angular, JavaScript, CSS, HTML)
• 5+ years working with the AWS cloud
• 1+ years working with the Azure cloud
• Familiarity with continuous integration/delivery practices
• Docker, Kubernetes experience a plus

Responsibilities

• Experience building agentic harnesses from scratch, including capabilities such as tool use, multi-step chaining, reasoning, streaming, skills, multimodal integration, RAG, sandboxing, and state management.
• Experience designing, building, and operating production APIs and services, including RESTful APIs, streaming APIs, asynchronous workflows, service boundaries, versioning, authentication/authorization, error handling, and backward compatibility.
• Strong judgment around when to use synchronous APIs, event-driven architectures, queues, background jobs, or streaming protocols based on latency, reliability, scalability, and user experience requirements.
• Demonstrated ownership of production systems, including process improvements, roadmap contributions, defect resolution, uptime and availability monitoring, and operational reliability.
• Strong backend engineering fundamentals, with comfort working across service design, APIs, batch workflows, orchestration, observability, and production support.
• Familiarity with evaluation techniques such as creating and maintaining evaluation datasets, running offline regression evals, monitoring online production performance, rubric-based scoring, self-verification, human-in-the-loop review, AI-as-judge methods, and quality/reliability analysis at scale.
• Clear understanding of foundational machine learning and statistical concepts, including sampling, statistical significance, overfitting/underfitting, precision and recall, and quality tradeoff analysis.
• Experience building and orchestrating large-scale batch workflows that use foundation models, including LLMs and VLMs, as well as pretrained open-source models and deep learning models.
• Ability to design systems that safely and reliably automate meaningful enterprise workflows using non-deterministic AI components.
• Strong judgment around reliability, failure handling, observability, human review, and operational safety.
• Deep understanding of how to make systematic tradeoffs between quality, reliability, latency, cost, explainability, and user experience in complex AI systems.
• Ability to design pragmatic architectures that balance innovation with production readiness.
• Comfort working in environments where model behavior is non-deterministic and system design must account for uncertainty, evaluation, monitoring, and graceful failure.
• Experience with both lexical and embedding-based search methods.
• Ability to reason about relevance, ranking, latency, recall, precision, indexing strategy, and retrieval performance.
• Experience working with foundation models and open-source LLMs beyond simple API calls.
• Familiarity with lower-level model behaviors and controls, including temperature, top-p sampling, logprobs, confidence scoring, prompting strategies, and model selection tradeoffs.
• Comfort and willingness to build front-end experiences using JavaScript/TypeScript and frameworks such as React.
• While the role is primarily backend and AI systems focused, the ability to contribute to user-facing product experiences is important.
• Willingness to work across the stack and own the end-to-end product experience is essential.
• Familiarity with tools such as FFmpeg, OpenCV, or similar media-processing libraries is a plus.

Benefits

• All your information will be kept confidential according to EEO guidelines.
• During employment, employees are treated without regard to race, color, religion, sex, national origin, age, marital or veteran status, medical condition or handicap, or any other legally protected status.
• At times, government agencies require periodic reports from employers on the sex, ethnicity, handicap, veteran and other protected status of employees. The purpose of this Administrative EEO Record is for statistical analysis only and is used to comply with government record keeping, reporting, and other legal requirements. Periodic reports are made to the government on the following information. The completion of the Administrative EEO record is optional. If you choose to volunteer the requested information, please note that all Administrative EEO Records are kept in a Confidential File and are not part of your Application for Employment or Personnel file.
• Please note: YOUR COOPERATION IS VOLUNTARY. INCLUSION OR EXCLUSION OF ANY DATA WILL NOT AFFECT ANY EMPLOYMENT DECISION.

What This Job Offers

• Job Type: Full-time
• Career Level: Principal
• Education Level: Associate degree