General Dynamics Mission Systems Projects Millions in Savings with Snowflake AI
General Dynamics Mission Systems uses Snowflake to power multiple AI use cases, including a reimagined supply chain workflow that’s expected to yield over $1.3 million in annual productivity savings.
$1.3M+Expected annual productivity savings from improved sales quotation processing
120xUp to 120x faster workflow to extract and save data from complex sales quotations
Industry
Public SectorLocation
Chantilly, VirginiaMission-critical solutions for the heroes who lead, serve and protect our world
From helping sailors navigate dangerous waters to giving downed pilots a lifeline to call home, General Dynamics Mission Systems (GDMS) provides solutions for those who lead, serve and protect. GDMS’ products, programs and services span across land, sea, air, space and cyber and range from spaceborne communication technologies to underwater sonar systems.
Developing mission-critical solutions for the world’s largest defense forces requires continual innovation that’s fueled by reliable, actionable data. GDMS uses the Snowflake AI Data Cloud to accelerate innovation, streamline business operations and keep employees focused on smarter, more secure missions.
Because we’re in a government environment, we needed a partner that could help us securely solve our ML and LLM challenges. Snowflake was the big solution.”
Obaid Masih
Senior Principal Data Scientist/MLOps Engineer, General Dynamics Mission Systems
Story highlights
Streamlined sales quotation workflow reduces costly data entry and delays: GDMS’ revamped workflow, powered by Snowflake, runs up to 120 times faster than before, which elevates productivity and mitigates friction with vendors.
Data scientists spend less time dealing with ML technical debt: Troubleshooting library dependencies has mostly become a thing of the past, leaving more time for innovation.
Interest in AI continues to grow among GDMS’ business users: From navigating subpoena requests to enabling intelligent data interaction via AI agents, GDMS’ AI project pipeline is robust.
Solving business challenges with LLMs while ensuring regulatory compliance
Successfully fulfilling a multimillion-dollar defense contract is a compliance-intensive, logistically challenging endeavor. One government contract could consist of hundreds of pages of clauses that GDMS must carefully match to relevant citations, terms and conditions. GDMS works with approximately 30,000 vendors, which introduces complexity, especially when it comes to managing the volume and variability of incoming sales quotations.
An upstart data science team began building solutions to alleviate regulatory and supply chain issues, but GDMS’ legacy data architecture posed difficulties. Internally sourced large language models (LLMs) delivered questionable accuracy and lacked document parsing capabilities. Maintaining ML workflows and troubleshooting library dependencies diverted the team from higher-impact initiatives. “Somehow we made it work and customers were happy, but it was a lot of effort for our very small team,” says Obaid Masih, Senior Principal Data Scientist/MLOps Engineer at General Dynamics Mission Systems.
Responding to business users’ need for enhanced sales quotation processing, Masih evaluated multiple platforms before selecting Snowflake. In addition to Snowflake’s support for customers with Federal Risk and Authorization Management Program (FedRAMP) compliance requirements, Snowflake Cortex AI was a key factor in the decision-making process. “An LLM is only as good as its scanning and parsing,” Masih says. “That’s why Snowflake became such a clear choice.” Snowflake Cortex AI Functions make it possible to run unstructured analytics on PDFs with industry-leading LLMs. “Having AI_PARSE_DOCUMENT and AI_COMPLETE was a game changer for us,” says Parvathi Susarla, Data Scientist at General Dynamics Mission Systems.
What I love about Snowflake Cortex is the AI_COMPLETE function. The output is standardized, meaning I can change the LLM quite easily.”
Obaid Masih
Senior Principal Data Scientist/MLOps Engineer, General Dynamics Mission Systems
Unlocking millions of dollars in productivity savings and healthier supplier relationships
GDMS anticipates over $1.3 million in annual productivity savings from its reimagined approach to sales quotation processing. “As more people use it, more savings will happen,” Masih says. “We’re expecting millions of dollars, nothing short of that.”
GDMS’ solution uses Snowflake Cortex AI as well as Snowflake Streams and Tasks to transform an operational bottleneck into a highly automated workflow. Extracting points of interest and writing outputs to Salesforce occurs within 2 minutes of quotation upload. For complex quotations, that’s up to 120 times faster than before.
It’s a huge success because it reduces manual processes and the time required to upload quotations. The time savings converts into dollar savings.”
Parvathi Susarla
Data Scientist, General Dynamics Mission Systems
Buyers perform considerably less manual data entry, and GDMS’ vendors are happier too. According to Masih, “They know their quotations will be processed faster.”
Avoiding unnecessary software licensing costs and optimizing machine learning operations (MLOps)
With Snowflake, GDMS’ data scientists spend less time deciphering ML dependencies and more time on value-added projects. For example, GDMS now ingests software licensing and usage data into Snowflake to power spend and usage analytics. “It reveals the software licenses you need and those you don’t,” Masih says. “It’s actually a very big money saver.”
Migrating additional use cases to Snowflake helps GDMS pay down technical debt and increase data’s impact. ML projects such as anomaly detection, tax and expense code matching, and headcount forecasting will soon live in Snowflake.
Developing AI/ML solutions to further expand operational efficiency and data access
Success with the AI Data Cloud has led to growing demand for data science at GDMS. “People saw how well we did with the sales quotations, and now they want us to solve their problems,” Masih says.
Using AI to offset legal expenses is a top priority. According to Susarla, “We’re talking millions of dollars in outside counsel hours.” GDMS’ legal discovery project will use Snowflake Cortex AI Functions to understand subpoena requirements and identify relevant items from thousands of documents. Another AI project planned for Snowflake could simplify contract intelligence while reducing dependency on outside legal counsel. “The legal contract project is very important for defense teams because everything we do is highly scrutinized by governments,” Masih says. “It’s a compliance, cost-saving and technology story.”
Snowflake’s AI and ML features align with the team’s vision of empowering business users with AI agents. “We have things like Snowflake CoCo and Snowflake Cortex AI Functions available to us,” Masih says. “That’s why Snowflake is a good match for us.”
