US Commercial Lender|Finance
Governed AI Agents for Lending Cut Credit Decision Time by 58%
The client, a US small business lender, reviewed 1,500 loan files a month by hand. Our governed AI agents cut credit decision time by 58%.
Read MoreClient name withheld under NDA. To honour our non-disclosure agreement and compliance obligations, we describe this client by industry and region only. The project scope, technology and outcomes are shared without details that could identify them.
The client is a US online stamp auction platform that runs on SQL Server and its infrastructure. It wanted AI search and recommendations while keeping that infrastructure in place.
The legacy system lacked natural language search and personalized suggestions for bidders. Its data came from large, varied sources, including bidding history, lot descriptions and metadata.
A 4-developer Bacancy Technology team built a Python AI engine that connects through FastAPI without disrupting the existing platform. Engineers cleaned the data and trained semantic search and recommendation models with OpenAI and HuggingFace embeddings. Collectors now search in natural language and get lot and bidding suggestions based on their activity. The modular engine can support image search later.
Our AI experts added modern search to a legacy platform without touching its core.
New AI features had to run alongside the existing SQL Server and the client infrastructure.
Solution: Our AI experts built a lightweight Python AI engine and exposed it through FastAPI REST endpoints. The platform gained modern features with no disruption to existing operations.
Bidding history, lot descriptions and metadata came in many formats.
Solution: Our data engineers cleaned the bidding history, lot descriptions and metadata. They used OpenAI and Hugging Face embeddings to train semantic search and recommendation models that work across inputs.
Collectors needed relevant lots and bidding suggestions instead of generic results.
Solution: Our team built AI algorithms that analyze user activity and preferences and return relevant lots and bids in real time. The models keep learning from user behavior.
The client wanted to add image search in a later phase.
Solution: Our engineers made the AI engine modular, so it can accept image embeddings from tools such as Google Vision and OpenCV in future releases.
Ready to start?
Get access to an experienced team of developers and engineers from Bacancy Technology, handpicked to ace your goals. Kickstart within 48 hours, no-risk trial.
Book A 30 Min CallUS Commercial Lender|Finance
The client, a US small business lender, reviewed 1,500 loan files a month by hand. Our governed AI agents cut credit decision time by 58%.
Read More
UK Precision Components Manufacturer|Manufacturing
The client technicians searched scattered manuals during breakdowns. Our Azure RAG platform cut downtime 30% and returns cited answers in seconds.
Read More
US Multi-Specialty Healthcare Group|Healthcare
The client, a 14-clinic US group, keyed prior auths into payer portals by hand. Our AI agent cut turnaround from 5 days to 1.5 days.
Read More