Generative AI Proof of Concept - Ideation to PoC

Engaging Model

• Define PoC requirements and collaborate with you to define KPIs and success criteria.
• Validate the GenerativeAI use case and business needs with your team.
• Analyze data availability, data sources, and quality, model training infrastructure readiness, model training, and deployment.
• Set up a proof-of-concept infrastructure and environment for model training and deployment.
• Determine and use applicable foundation models based on use cases and data availability, such as GPT 3.5/4.0, Llama 2, etc.
• Use a vector database to load your company's specific data for in-context prompting (optional).
• Fine-tuning the model and prompt optimization to a given use case.
• PoC can last up to 8 weeks, divided into two phases: readiness (1 week) and execution (7 weeks).
• This Proof of Concept will cover up to one Generative AI use case.
• We will collaborate with AWS to secure the funds required to cover the cost of the PoC services.

Deliverables

• Working PoC and Demo
• Documented problem statement, Generative AI use case, and KPI and success criteria
• High-level PoC design and technical documentation
• All the software code, prompts, and fine-tuning developed for the PoC
• Documented roadmap to scale and move the PoC to MVP

Highlights

• Working prototype for the use case to demonstrate the value to business
• MVP roadmap and execution strategy for the next phase
• ROI and business use case value justification

‍