Measure an existing LLM workflow against real examples, identify weak points, and implement targeted improvements with honest before-and-after evidence.
WHAT I WILL BUILD
I will evaluate one existing LLM application or workflow using a defined set of representative cases and agreed success criteria. First, I will establish an honest baseline from the current implementation. I will then identify the highest-value issues within the available evidence and implement targeted improvements, which may include prompt changes, retrieval adjustments, output validation, or workflow refinements as appropriate to the application. I will report the before-and-after evaluation results for the agreed cases, including limitations and cases that remain weak. Improvement is not guaranteed: results depend on the application, model, data, and evaluation set. The handover includes updated source/configuration where changes are made, the evaluation cases or format, and a concise findings summary. One revision round is included within the agreed application and evaluation brief. You supply authorized access, a runnable baseline, representative test cases or the ability to define them together, and relevant credentials. Hosting, model, API, cloud, and third-party software fees are separate.
INCLUDED IN THIS FIXED-PRICE PACKAGE
• Evaluation plan with agreed success criteria and representative cases
• Documented baseline results from the existing implementation
• Targeted implementation changes for agreed issues within one application
• Before-and-after results for the agreed evaluation set, with limitations noted
• Updated source/configuration and evaluation handover materials
WHAT I NEED FROM YOU
• Authorized access to a runnable existing LLM application
• Current prompts, retrieval configuration, or workflow documentation where available
• Representative cases, expected outcomes, and rules for judging results
• A stakeholder who can confirm business priorities and review findings
• Required model, database, and hosting credentials
SCOPE BOUNDARIES
• Building a new LLM application from scratch
• Guaranteed quality gains, accuracy targets, or production outcomes
• Open-ended evaluation of every possible user input or model behavior
• Broad platform rearchitecture or multiple unrelated application workflows
• Ongoing monitoring, hosting, model/API, cloud, and third-party fees
DELIVERY AND NEXT STEP
Fixed-price starter package: USD 950. Delivery: 10 days after the brief, authorized access and required materials are agreed and provided. One consolidated revision round within the agreed scope. Contact me before ordering so we can confirm compatibility, integration boundaries and acceptance criteria. Hosting, hardware, model/API usage and third-party fees are separate.
