LankaBangla: Air-Gapped Market AI

LankaBangla Securities is one of Bangladesh's leading capital-market institutions, with rich client and market data — and, until this engagement, no way to reach it without a technical intermediary.

The challenge

Dealer-brokers needed synthesized answers inside decision windows, not ad-hoc reports days later. Retail outreach needed relevance grounded in account and trading context, not generic campaigns. And in a regulated institution, free-form AI access to production data was never an option — any system had to be read-only, role-aware, and fully auditable by design.

Our approach

Architecture first: isolate the model in an air-gapped, open-source LLM environment, then mediate every interaction through a governed application layer. Natural-language requests resolve to pre-approved, read-only query templates — the model never touches production data directly, and every interaction is logged.

What we built

Two working functions on one reusable foundation: an AI assistant that lets dealer-brokers retrieve and synthesize information in seconds, and marketing automation for retail traders that adapts messaging to trading behavior and account context. Because both run on the same governed layer, new functions can be added without re-architecting the platform.

The outcome

The institution gained practical AI adoption without regulatory exposure: conversational access for the people who need speed, controlled execution for the people who answer for it, and an extensible base for what comes next.