OptaAI — an AI-native sourcing platform
OptaAI is an AI-native sourcing and procurement platform we built for Optamark, a US promotional-products company, to replace manual supplier-matching with semantic, natural-language search — running on Spring AI, OpenAI embeddings and Elasticsearch vector search.
The challenge
Sourcing across a large supplier and product catalogue was slow and leaned on tribal knowledge. Buyers and sales reps needed to find the right product or supplier from a natural-language description — not an exact keyword match — and the manual workflow couldn't keep up.
What we built
A semantic search and AI-assistant layer over the catalogue. Products and documents are embedded with OpenAI embedding models and stored in an Elasticsearch vector index. Spring AI's ChatClient and VectorStore abstractions drive retrieval-augmented answers, so results are grounded in the real catalogue rather than invented by the model.
Zero-downtime reindexing
The catalogue and its embeddings change constantly. We used a rolling-index strategy behind an alias, so re-embedding and reindexing happen with no search downtime and no stale or partial results during cutover — search stays live and correct throughout.
An MCP server for reps and self-serve
We exposed the sourcing capabilities through a Model Context Protocol (MCP) server, so the same tools power both an internal assistant for sales reps and customer self-serve. Building a native MCP server in Java is a genuinely leading-edge pattern, and it let one well-tested toolset serve multiple front ends.
Production hardening
Token budgeting, caching, timeouts with graceful fallback, and evaluation against a fixed test set before each release — the work that separates a demo from a system a business runs on every day.
Outcomes
- Client-reported ~3× sales-team productivity on sourcing tasks.
- Client-reported ~70% faster query response versus the prior manual workflow.
- Semantic, natural-language search over the full catalogue with grounded, citable answers.
- An ongoing, multi-year partnership.
Figures are client-reported or from internal benchmarks and are illustrative of typical results, not independently audited.
Want this kind of result on your platform? Explore our Spring AI services, see how we work with overseas clients, or talk to an architect.