Finance ops · Production

MNP builds a RAG-based GenAI client insights platform on Databricks in under six weeks

The problem

MNP's existing systems could not support the processing required for advanced ML applications, and their initial GenAI deployment faced integration constraints, response latency, and prohibitive GPU costs that made scaling a client insights solution impossible.

First attempt

MNP's initial GenAI deployment was blocked by its foundation model's tight coupling to the existing data warehouse, time-to-first-token latency issues, and GPU costs that became prohibitive at the scale needed for a reliable corpus.

Workflow diagram · grounded in source
1
Client submits business query
trigger
“an intuitive, language-based interface for clients to make insightful, data-driven decisions about their businesses and industries”
2
Data consolidated into lakehouse
integration
“The Databricks Data Intelligence Platform served as a unifying force to consolidate MNP's structured, semi-structured and unstructured data into a single repository”
3
Vector Search retrieves embeddings
ai_action
“help the firm's clients and client-facing teams pinpoint the most relevant information for specific queries”
4
RAG integrates retrieved context
ai_action
“The ability to continually integrate new data, refresh the embedding database and employ the information to provide contextual relevance was fundamental in MNP's strategic initiative to deploy RAG within the Databricks Data Intelligence …”
5
LLM generates summarized insights
ai_action
“MNP recognized the potential of LLMs and their summarization capabilities and wanted to offer an intuitive, language-based interface for clients to make insightful, data-driven decisions about their businesses and industries”
6
Insights delivered to client
output
“Databricks Model Serving was also implemented to ensure that the models were constantly available and responsive to queries”
Reported outcome

MNP developed a GenAI solution delivering accurate responses in less than six weeks, with the data team building a model from start to quality assurance testing within four weeks, while maintaining data isolation via Private AI standards.

Reported metrics
GenAI solution development timeless than six weeks
model build to QA testing timewithin four weeks
Reported stack
DatabricksVector SearchMixtral 8x7BRAGUnity CatalogDatabricks Foundation Model APIsDatabricks Model ServingLlama 2 13BDatabricks GenAI Advisory Program
Source
https://www.databricks.com/customers/mnp
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

MNP developed a GenAI solution delivering accurate responses in less than six weeks, with the data team building a model from start to quality assurance testing within four weeks, while maintaining data isolation via…

What tools did this team use?

Databricks, Vector Search, Mixtral 8x7B, RAG, Unity Catalog, Databricks Foundation Model APIs, Databricks Model Serving, Llama 2 13B, Databricks GenAI Advisory Program.

What results were reported?

GenAI solution development time: less than six weeks; model build to QA testing time: within four weeks (source-reported, not independently verified).

What failed first in this deployment?

MNP's initial GenAI deployment was blocked by its foundation model's tight coupling to the existing data warehouse, time-to-first-token latency issues, and GPU costs that became prohibitive at the scale needed for a r…

How is this finance ops AI workflow structured?

Client submits business query → Data consolidated into lakehouse → Vector Search retrieves embeddings → RAG integrates retrieved context → LLM generates summarized insights → Insights delivered to client.