Muhammad Bilal Tahir

Challenge — GL reconciliation and ad-hoc reporting required an engineer in the loop for every new question stakeholders had, slowing decisions down.

Approach — Architected an agentic workflow on Azure AI Foundry so stakeholders can ask questions in plain language; the agent plans and executes the query safely, automates GL parsing and matching, and cites its sources. Paired it with a Qdrant-backed retrieval layer for grounded, contextual answers, plus dashboards for chat and history tracking.

Impact — Cut reporting and reconciliation turnaround by 60–70%, taking a multi-day manual process down to a few hours; adopted by multiple analyst and finance teams across the organization for day-to-day reporting.

Challenge — The client needed to give account holders a real-time, unified view of their finances across multiple banks, without building banking infrastructure themselves.

Approach — Integrated the Plaid API to aggregate live financial data from partner banks, built token-based authentication and encryption for secure data handling, and shipped a dynamic Angular front-end backed by Identity Server for authorization.

Impact — Connects accounts across major US banks via Plaid for real-time identity and account verification, on a compliance-first, SOC 2-ready architecture with encryption at rest and in transit.

Challenge — A monolithic firmware-update system couldn't scale with a growing device fleet, risking slow, unreliable rollouts of critical updates.

Approach — Led the migration from a monolithic to a microservices architecture, integrated LWM2M for lightweight device-to-server communication, and optimized PostgreSQL to handle high-volume transactional load under strict OTA security compliance.

Impact — Manages OTA updates across a fleet of thousands of connected IoT devices, with a 99%+ update success rate driven by staged rollouts and automated rollback — supporting millions of device transactions in production.