Design Applications

A customer support team receives tickets in multiple languages. The GenAI application must: (1) detect and translate each ticket into English, (2) classify the ticket into a support category, and (3) generate a concise English summary for the support agent. The team wants a maintainable design where each step can be tested and swapped independently, and where a failure or low-confidence result in one step can be inspected before the next step runs. Which design approach best meets these requirements?

A software company wants to build a generative AI assistant that reviews pull requests. For each PR, the assistant must: (1) summarize the code changes, (2) flag potential security vulnerabilities against the company's internal secure-coding guidelines, and (3) generate a natural-language review comment that references specific findings. The security guidelines are a large, frequently updated internal document. As the GenAI engineer designing this application, what is the most appropriate high-level architecture?

A financial services company wants to build a customer support assistant that can answer questions about a customer's account. A single request may require the assistant to: (1) look up the customer's current account balance from an internal transactional database, (2) retrieve relevant policy explanations from an internal knowledge base of PDFs, and (3) perform a currency conversion calculation before presenting the final answer. The team wants a design where the LLM dynamically decides which capability to invoke based on the user's question. Which application design best fits these requirements?

A financial services firm wants to build a system that answers analyst questions like 'Summarize the risk factors from the three latest 10-K filings for Company X and compare them to sector peers.' The number of documents to retrieve and the comparison steps vary per query, and the system sometimes needs to call a live market-data API mid-workflow depending on what the question asks. The team must decide how to structure the application. Which design approach best fits these requirements?

You are designing a customer-support generative AI application for an e-commerce company. Business requirements state that the assistant must: (1) answer general product policy questions, (2) look up a specific customer's live order status by order ID, and (3) initiate refund requests that require writing to an internal transactional system. The company wants a single conversational entry point that decides which capability to use per turn. Which application design best satisfies these requirements?

A retail company wants a generative AI application that lets customer support agents upload a photo of a damaged product and receive a plain-language description of the visible damage, which is then combined with the customer's typed complaint to draft a suggested resolution email. The engineering team must translate this business requirement into an application design. Which design approach best satisfies the requirement?

A Generative AI Engineer is designing a customer-support RAG assistant for a mid-size retailer. Business requirements state that roughly 80% of incoming questions are simple FAQ-style lookups (store hours, return policy) that must be answered in under 1 second at low cost, while the remaining 20% are complex, multi-part troubleshooting questions where answer quality matters most and slightly higher latency is acceptable. The team has both a small, fast, inexpensive LLM and a large, high-quality, more expensive LLM available as Model Serving endpoints. Which application design best satisfies these requirements?

A financial services team is building an assistant that must solve multi-step budgeting calculations submitted by clients (e.g., 'If I save 12% of my $4,200 monthly income for 18 months, then withdraw 30%, how much remains?'). Early tests with a strong instruction-tuned LLM show the final numeric answer is frequently wrong, even though the model clearly understands the question. The engineer wants a prompt strategy to improve arithmetic accuracy without fine-tuning or adding external tools yet. Which prompt strategy should the engineer apply first?

A financial services company is building a customer support triage assistant. Incoming support emails must be classified into one of five predefined categories (e.g., 'billing', 'fraud', 'account_access'), and the output must be returned as structured JSON so a downstream workflow can route the ticket automatically. The team has no labeled training data but has a clear written description of each category. Which prompt strategy best fits this use case?

A retail company is building a customer-facing chatbot to respond to product inquiries. The business requirement states that every reply must maintain a warm, empathetic brand voice and consistently structure responses as a short greeting followed by the answer and a closing offer to help further. The team is using a general-purpose instruction-tuned LLM without fine-tuning access. During testing, responses vary widely in tone and format. Which prompt strategy best addresses this requirement with minimal engineering effort?

A legal-tech team is building a RAG application to answer questions about lengthy commercial contracts. The documents contain deeply nested clauses, cross-references between sections, and formal defined terms that only make sense within their surrounding context. Early testing shows the assistant frequently returns partial answers that omit conditions stated in adjacent clauses. Which chunking strategy should the engineer adopt to best preserve semantic coherence for retrieval?

A Generative AI Engineer at an insurance company is building a RAG application that answers agent questions using a corpus of 50,000 policy documents stored as Delta tables in Unity Catalog. The documents are updated frequently as policies change, and the team wants the retrieval layer to automatically stay in sync with the source Delta table without manual re-embedding jobs. They also want to reuse Databricks-managed governance and lineage. Which retrieval component design best satisfies these requirements?

A pharmaceutical company is building a RAG application over a corpus of internal drug-research documents that are dense with specialized biomedical terminology and abbreviations. During evaluation, the team notices that retrieval frequently returns loosely related passages and misses the technically correct sections, even though the vector index and chunking are configured correctly. As the Generative AI Engineer, which change to the RAG design is MOST likely to improve retrieval relevance?

A financial services company is building a RAG-based internal assistant that answers employee questions about corporate HR policies. Compliance requires that the assistant NEVER fabricate policy details and must clearly refuse when a question falls outside the indexed policy documents. During testing, the LLM sometimes confidently answers questions about topics not present in the retrieved context. Which design change most directly addresses this requirement?

A Generative AI Engineer is building a customer-support chatbot on Databricks that answers questions from a product documentation knowledge base. Users frequently ask follow-up questions such as 'What about the enterprise tier?' that only make sense in the context of earlier turns in the conversation. Testing shows that when these follow-up questions are sent directly to the vector search retriever, the retrieved chunks are irrelevant because the standalone query lacks context. Which design change best addresses this problem while keeping retrieval accurate?

A generative AI engineer is building a RAG-based internal knowledge assistant. During evaluation, the team finds that the vector search step retrieves 20 chunks that contain the relevant answer somewhere in the set, but the chunks actually passed to the LLM (top 3 by vector similarity) frequently miss the most relevant passage, causing incomplete answers. The embedding model and chunking strategy are already well-tuned. Which change to the retrieval architecture would most directly improve the relevance of the final chunks sent to the LLM?

A Generative AI Engineer is building a RAG application over a corpus of long internal policy documents. Each document covers multiple unrelated topics (e.g., a single HR handbook covers benefits, leave policy, and code of conduct). During testing, retrieval frequently returns chunks that mix relevant and irrelevant content, and the LLM produces answers that blend unrelated topics. Which change to the retrieval architecture will most directly improve the relevance of retrieved context?

An insurance company wants a chatbot that answers employee questions about internal policies. The policy documents change several times per month, and answers must always cite the specific current policy section. The engineering team must recommend an application design approach. Which approach best satisfies these business requirements?

A retail company wants to generate personalized product description summaries for all 500,000 items in their catalog. The summaries will be displayed on product pages but only need to be refreshed once per week when the catalog is updated. The engineering team is deciding how to design the generative AI serving approach. Which design best fits these business requirements?

A healthcare provider wants to build a patient-facing chatbot that answers questions about appointment scheduling and general wellness. Legal has mandated that the application must never expose or store any protected health information (PHI) and must refuse to provide specific medical diagnoses. As the GenAI engineer, which design approach best translates these business requirements into the application architecture?