> SYSTEM_READY
Abdul Rahim Moinul Haque.
Full Stack AI Engineer. Architecting Intelligence.
I architect high-performance, real-time AI systems and enterprise-grade backend microservices. From sub-second latency voice models and multi-modal WebRTC pipelines to robust computer vision tracking and scalable omnichannel conversational agents.
Technical Competencies
Core AI & Machine Learning
- ▹LLM Orchestration & Prompt Engineering
- ▹Computer Vision (ONNXRuntime, InsightFace)
- ▹Multimodal AI Integration (Vision, Audio)
- ▹Retrieval-Augmented Generation (RAG)
- ▹Voice Activity Detection (VAD)
- ▹STT/TTS Streaming Pipelines
- ▹Zero-Shot Data Augmentation
Backend & Infrastructure
- ▹Python & FastAPI
- ▹Node.js
- ▹Live Video Streaming (RTSP, HLS, FFmpeg)
- ▹WebRTC (LiveKit)
- ▹SIP Telephony (Twilio)
- ▹Vector Databases (Pinecone, Redis)
- ▹Event-Driven Architecture
Frontend & Integration
- ▹React.js & Next.js
- ▹Tailwind CSS
- ▹State Management & Real-time UI
- ▹MCP Tool Management
- ▹Omnichannel APIs (WhatsApp, Meta Graph)
- ▹Mobile Web Push Notifications
Engineering & Collaboration
- ▹System Architecture Design
- ▹High-Traffic Scaling Optimization
- ▹Sub-second Latency Profiling
- ▹Hot-Reloading & Zero Downtime Systems
- ▹Cross-Functional Team Leadership
- ▹Automated CI/CD Pipelines
Work Experience
AI Engineer
May 2025 - Present- ▹Architected an Omnichannel AI Agent to handle massive traffic, utilizing MCP tool management and integrating vector databases (Pinecone, Redis) for semantic search and persistent context. Connected to official WhatsApp and Meta Graph APIs for business automation.
- ▹Designed a sub-second latency Real-Time AI Voice Platform bridging SIP telephony via Twilio and local FreePBX/GSM with web clients. Leveraged OpenAI Realtime, Gemini Live, XAI Voice API, Deepgram, and LiveKit with sophisticated tool calling and Mem0 memory orchestration.
- ▹Developed a multi-modal LiveKit WebRTC platform bridging interactive 3D avatars with remote LLM backends, featuring precise cognitive state synchronization and zero REST bottlenecks.
- ▹Engineered a zero-downtime RTSP Computer Vision Pipeline executing continuous face alignment and extraction on CPU-optimized ONNX models (YuNet, InsightFace). Features hot-swappable in-memory vector indices and real-time mobile push notifications.
Backend Developer
Jan 2024 - May 2025- ▹Architected and maintained secure, enterprise-grade backend microservices using .NET Framework and C#, optimizing API performance and cross-service communication via REST and gRPC.
- ▹Engineered complex SQL queries and robust database architectures within PostgreSQL and MS SQL Server, significantly reducing query latency for high-throughput data retrieval.
- ▹Containerized backend applications using Docker and orchestrated deployments via Kubernetes to ensure high availability, automatic scaling, and resilient failover.
- ▹Streamlined continuous integration and deployment lifecycles by configuring automated CI/CD pipelines, integrating rigorous unit testing and secure artifact management.
- ▹Implemented strict OAuth2.0 and JWT-based authentication flows, ensuring secure data transit and stringent access control across all internal and external facing endpoints.
Featured Projects
OMNICHANNEL AI AGENT
An advanced AI Chat agent built with MCP tool management and memory context, integrating vector databases for robust semantic search. Engineered to handle massive traffic while maintaining individual chat memory to recall previous context. Seamlessly integrated with WhatsApp and social media platforms through official APIs to help businesses automate lead qualification and streamline customer inquiries at scale.
Key Engineering
- ▹MCP Tool Management for dynamic agent workflow execution
- ▹Vector Database Integration for semantic memory recall across sessions
- ▹High-Traffic Architecture optimized for asynchronous messaging
- ▹Omnichannel APIs binding WhatsApp, Meta Graph, and internal CRMs
AI VOICE
A low-latency, real-time conversational AI platform bridging SIP telephony and web clients with advanced multimodal LLMs. Orchestrates custom STT/TTS streaming pipelines to deliver human-like voice interactions with robust state management, dynamic workflow tool execution, and persistent contextual memory. Features a local UAE GSM integration via FreePBX to enable seamless inbound local calls to the AI agents.
Key Engineering
- ▹Real-Time AI Audio streams via OpenAI Realtime, Gemini Live, and XAI Voice API
- ▹Sophisticated Tool Calling orchestration across diverse multimodal LLMs
- ▹Advanced State Orchestration with VAD-triggered interruptions
- ▹Telephony AI Bridge connecting traditional SIP networks via Twilio and local GSM/FreePBX infrastructure
- ▹Post-Call AI Analytics for automated transcription and intent extraction
AI AVATAR
Architected a real-time WebRTC platform that brings AI avatars to life with sub-second latency voice and video streaming. By bridging advanced AI agents with a robust LiveKit infrastructure, the system orchestrates dynamic conversation states (listening/thinking/speaking) and multi-modal inputs to deliver a seamless, human-like interactive experience.
Key Engineering
- ▹Sub-Second Latency Streaming bypassing traditional REST bottlenecks
- ▹Cognitive State Synchronization mapping AI processing to UI feedback
- ▹Dynamic AI Provisioning matching visual avatars with specialized LLMs
- ▹Multi-Modal Pipelines routing local camera feeds to vision models
ATTENDANCE SYSTEM
A real-time AI identity and tracking pipeline that processes live RTSP camera feeds to autonomously recognize and log human presence. Engineered as a unified, single-server architecture, it seamlessly combines a low-latency computer vision inference loop with a full-stack Node.js and React application to provide sub-second facial recognition, hot-reloading vector indices, and instant push notifications.
Key Engineering
- ▹Live Vision Pipeline executing face alignment and 512-dim vector extraction on CPU-optimized ONNX models
- ▹Hot-Swappable AI State with an in-memory cosine similarity matrix that rebuilds with zero downtime
- ▹Automated Data Augmentation generating robust vector variations from single user registration photos
- ▹Instant AI Event Hooks tied directly to mobile Web Push pipelines for real-time entry/exit alerts