Abdul Rahim
Moinul Haque

Own real-time AI and agentic systems end-to-end — from voice, avatar and vision infrastructure to multi-agent LLM platforms — and take them to production.

• Built an omnichannel customer chat platform across WhatsApp, Instagram and web, where an AI agent handles insurance enquiries and submissions and staff can step in at any time — with safeguards preventing unverified submissions, live human handover over Socket.IO, and encryption for customer data. The agent runs on Python/Flask with PostgreSQL, the live console on Node/Express with MongoDB, using OpenAI function calling, voice notes and OCR.
• Developed a platform that lets clinicians research, write and fact-check medical content and produce videos in a clone of their own voice. It runs on a five-agent team (Agno) with a workflow engine that recovers cleanly after a crash, routes each task to a lighter or heavier model based on complexity, and cites its sources with RAG — with multi-language support including Arabic right-to-left.
• Designed and built a real-time voice assistant on LiveKit linking phone lines (Twilio, plus on-site FreePBX/GSM) to the browser — running STT, the model and TTS fast enough to respond in under a second, with voice-activity detection and memory that persists across calls. With the voice layer stable, I layered a talking video avatar on top that stays in sync with the conversation over the same connection.
• Created a computer-vision system for a large multi-level car park that identifies the fastest exit for each vehicle and helps reduce congestion — using YOLO to detect vehicles and read number plates across several camera feeds, then scoring the candidate exit routes in real time.

Built and hardened enterprise-grade backend services and the delivery pipelines around them.

• Built and maintained backend microservices in C# and .NET, refining how they communicated over REST and gRPC to keep the APIs responsive.
• Wrote the SQL and designed the schemas in PostgreSQL and MS SQL Server, reducing query times on the heavier data-retrieval paths.
• Containerized the services with Docker and ran them on Kubernetes, so they scaled automatically and recovered from failures without downtime.
• Configured the CI/CD pipelines, including unit tests and a managed artifact store, so releases shipped consistently.
• Implemented authentication with OAuth2.0 and JWT, tightening access control across internal and public-facing endpoints.

Digitised how the business ran — building its CRM from scratch and leading the ERP selection and rollout.

• Built a custom CRM from scratch in Python, .NET and React, designed around how their sales and customer teams actually worked — bringing client data into one place and removing much of the manual effort from daily operations.
• Evaluated ERP options against their accounting and operational needs, selected Odoo, and managed its setup and rollout across finance and operations.
• Trained staff on the new CRM and ERP and produced the supporting documentation, making the move from their previous systems straightforward.