AI Engineer (III)

We need to output HTML fragment with sections: About the role (maybe overview), Responsibilities, Qualifications, Benefits (since Salary And Benefits present). Also maybe Pay? The posting includes "Competitive salary package" but no specific numbers, so we should not create Pay section because it's not specific. Benefits we have details: health, dental, vision, retirement savings plan with company match, paid time off and holidays, professional development opportunities. So we can include Benefits section. Schedule: "In this role you would be expected to come into the office 2-3 days a week (Tues & Wed are in office days)." That's schedule info. So we can include Schedule section. We should not repeat job title. Use

for each section heading. We need to preserve all substantive info. Remove boilerplate like company overview? The overview includes "Company is at the forefront..." That's substantive? It's about company, but we can keep as part of About the role? The instruction: "Prefer a few substantive sections over many thin ones." We can have About the role section containing the overview paragraphs (first two paragraphs). Then Responsibilities section with list items. Qualifications section with list. Benefits section with list. Schedule section with paragraph. We must not include metadata like Job ID. Not present. We must not include application instructions, EEO, etc. Not present. We need to use

• for any list of duties/requirements/benefits. For responsibilities we have many bullet-like sentences; we can turn each into
• . For qualifications we have list-like sentences; we can turn each into
• . For benefits we have list. We need to preserve all substantive info; do not invent or summarize away detail. So we need to keep details like specific frameworks, etc. Let's extract overview paragraphs: "Company is at the forefront of the Generative AI revolution, dedicated to shaping the future of artificial intelligence. Our Agentic AI team is focused on driving innovation by building next generation AI applications and enhancing existing systems with Generative AI capabilities. We are seeking an AI Engineer to help us with the development, deployment, and scaling of advanced AI applications that address real-world challenges. In this role, you will focus on designing, building, and scaling advanced AI applications in the PropTech space. You will work with state-of-the-art foundation models, RAG architectures, and multi-agent systems while partnering closely with product, design, and engineering teams. You will be responsible for taking AI concepts from ideation to production, owning end-to-end solutions that improve our products and transform user experiences. In this role you would be expected to come into the office 2-3 days a week (Tues & Wed are in office days)." We can split: first part about company and role overview (maybe About the role). Then schedule sentence separate. Responsibilities list: The responsibilities section heading appears as "Responsibilities". Then bullet points: "Agentic AI & Generative Application EngineeringEvaluate and use LLMs and multimodal models from multiple providers (e.g., OpenAI, Google, Anthropic, etc.) for:Conversational assistants, task-based copilots, and AI agentsSummarization, content generation, document understanding, generative analyticsBasic multimodal use cases (text + image, text + document, and soon video/audio)Design and implement agentic workflows (e.g., tool-calling, multi-step reasoning, multi-agent orchestration) using:LangChain, OpenAI Agents SDK, Google ADK or similar frameworks.Prompt Engineering & Guardrails Design and optimize prompts and system instructions to:Improve task completion, reliability, and latencyMinimize hallucinations and toxic/unsafe outputsImplement structured outputs (JSON/JSON Schema)Develop function/tool calling and prompts that help AI call them properlyIntegrate safety/guardrail layers (e.g., content moderation APIs, Guardrails AI, Rebuff, custom policies) to keep conversations focusedRAG & Knowledge IntegrationArchitect and implement RAG pipelines:Choose and configure vector databases (e.g., PGVector, Vertex AI Search, Pinecone, etc.)Build ingestion pipelines for internal data (documents, tickets, logs, property data, etc.)Implement knowledge retrieval process that draws from multiple sources and uses reranking to improve the response quality.Explore emerging retrieval techniques (semantic caching, knowledge graphs, long-context models, memory systems).Full-Stack & System IntegrationBuild or integrate front-end experiences (React / Vue / Svelte / Web RTC) for AI agents and copilots.Develop back-end services to orchestrate AI calls using REST, gRPC, WebSockets, or MCP; ensure scalability and observability.Integrate with internal systems and PropTech data sources using secure APIs and data contracts.Evaluation, Monitoring & OptimizationDesign and maintain evaluation pipelines and benchmarks for LLM-based features:Offline metrics (accuracy, relevance, latency, cost)Human-in-the-loop evaluations where neededUse AI observability and tracing tools (e.g., LangSmith, OpenTelemetry, etc.) to monitor quality.Optimize for performance, reliability, latency, and cost through:Model selection and routing (e.g., small vs. large models, Google vs. OpenAI)Prompt/token optimization and caching strategies.Collaboration, Documentation & DeliveryCollaborate with cross-functional teams (Product, Design, Domain Experts, Data Science, Platform Engineering) to define requirements and success metrics.Participate in architecture and design reviews; write clear technical documentation and runbooks.Contribute to shared libraries, templates, and best practices for AI development.Work in an Agile environment and own features from design through deployment and maintenance." We need to break into sensible list items. Could group by headings but we can just make each sentence as
• . However many are long run-on sentences without punctuation. We need to parse them. Let's split by periods maybe. I'll create list items: - Evaluate and use LLMs and multimodal models from multiple providers (e.g., OpenAI, Google, Anthropic, etc.) for: - Conversational assistants, task-based copilots, and AI agents - Summarization, content generation, document understanding, generative analytics - Basic multimodal use cases (text + image, text + document, and soon video/audio) - Design and implement agentic workflows (e.g., tool-calling, multi-step reasoning, multi-agent orchestration) using: - LangChain, OpenAI Agents SDK, Google ADK or similar frameworks - Prompt Engineering & Guardrails: Design and optimize prompts and system instructions to: - Improve task completion, reliability, and latency - Minimize hallucinations and toxic/unsafe outputs - Implement structured outputs (JSON/JSON Schema) - Develop function/tool calling and prompts that help AI call them properly - Integrate safety/guardrail layers (e.g., content moderation APIs, Guardrails AI, Rebuff, custom policies) to keep conversations focused - RAG & Knowledge Integration: Architect and implement RAG pipelines: - Choose and configure vector databases (e.g., PGVector, Vertex AI Search, Pinecone, etc.) - Build ingestion pipelines for internal data (documents, tickets, logs, property data, etc.) - Implement knowledge retrieval process that draws from multiple sources and uses reranking to improve the response quality - Explore emerging retrieval techniques (semantic caching, knowledge graphs, long-context models, memory systems) - Full-Stack & System Integration: - Build or integrate front-end experiences (React / Vue / Svelte / Web RTC) for AI agents and copilots - Develop back-end services to orchestrate AI calls using REST, gRPC, WebSockets, or MCP; ensure scalability and observability - Integrate with internal systems and PropTech data sources using secure APIs and data contracts - Evaluation, Monitoring & Optimization: - Design and maintain evaluation pipelines and benchmarks for LLM-based features: - Offline metrics (accuracy, relevance, latency, cost) - Human-in-the-loop evaluations where needed - Use AI observability and tracing tools (e.g., LangSmith, OpenTelemetry, etc.) to monitor quality - Optimize for performance, reliability, latency, and cost through: - Model selection and routing (e.g., small vs. large models, Google vs. OpenAI) - Prompt/token optimization and caching strategies - Collaboration, Documentation & Delivery: - Collaborate with cross-functional teams (Product, Design, Domain Experts, Data Science, Platform Engineering) to define requirements and success metrics - Participate in architecture and design reviews; write clear technical documentation and runbooks - Contribute to shared libraries, templates, and best practices for AI development - Work in an Agile environment and own features from design through deployment and maintenance That seems comprehensive. Qualifications section: extract: "Qualifications5+ years of total experience in Software Engineering and/or Data Science, with at least 2 years focused on Generative AI/LLMs.Degree in Computer Science, Machine Learning, Data Science, or related field, or equivalent practical experience.Strong proficiency in:Python for AI/ML, data pipelines, and back-end servicesJavaScript/TypeScript for front-end and/or Node servicesSQL and experience working with relational databases and basic data modelingWorking with coding assistants like Windsurf, Cursor, Codex, etc.Proven experience building production-grade software:Writing clean, testable, maintainable codeUsing CI/CD pipelines, code reviews, and Git workflowsHands-on experience with:At least one agentic/orchestration framework (OpenAI Agents SDK, Google ADK, LangChain, etc.)LLM APIs and/or open-source models (e.g., via OpenAI, Google, Hugging Face, Ollama)Vector embeddings, vector databases, and RAG architecturesExperience with one or more major cloud platforms (GCP, Azure, and AWS) and:Docker for containerizationKubernetes or a managed container service (e.g., EKS, GKE, AKS)Strong communication skills and ability to collaborate with both technical and non-technical stakeholders." We can break into list items: - 5+ years of total experience in Software Engineering and/or Data Science, with at least 2 years focused on Generative AI/LLMs - Degree in Computer Science, Machine Learning, Data Science, or related field, or equivalent practical experience - Strong proficiency in: - Python for AI/ML, data pipelines, and back-end services - JavaScript/TypeScript for front-end and/or Node services - SQL and experience working with relational databases and basic data modeling - Working with coding assistants like Windsurf, Cursor, Codex, etc. - Proven experience building production-grade software: - Writing clean, testable, maintainable code - Using CI/CD pipelines, code reviews, and Git workflows - Hands-on experience with: - At least one agentic/orchestration framework (OpenAI Agents SDK, Google ADK, LangChain, etc.) - LLM APIs and/or open-source models (e.g., via OpenAI, Google, Hugging Face, Ollama) - Vector embeddings, vector databases, and RAG architectures - Experience with one or more major cloud platforms (GCP, Azure, and AWS) and: - Docker for containerization - Kubernetes or a managed container service (e.g., EKS, GKE, AKS) - Strong communication skills and ability to collaborate with both technical and non-technical stakeholders Nice-to-Have skills we could optionally include but not required; we could omit as not a section unless we want to include as "Nice-to-Have Skills". The instruction: include a section ONLY when posting has real, specific content. It does have Nice-to-Have skills. Could we include a section? Possibly but we prefer few substantive sections. Might be okay to include as a section but it's not required. However we already have many sections; maybe we can omit Nice-to-Have to keep fewer sections. The instruction: "Prefer a few substantive sections over many thin ones." Nice-to-Have could be considered thin. I'd omit. Benefits: "Salary And Benefits Competitive salary package along with a comprehensive benefit plan that includes:Health, dental, and