I am an AI engineer and researcher with 7+ years of experience delivering production AI systems across Bangladesh, Japan, Saudi Arabia,USA, and the UK. My work spans LLM-powered chatbots and RAG systems, computer vision surveillance platforms, emotion detection engines, multi-agent CRM automations, and full-stack SaaS products. I currently serve as an Assistant Programmer at Bangladesh House Building Finance Corporation (BHBFC), developing FinTech modules for the government's home-loan platform.
My research focuses on the intersection of multimodal deep learning and healthcare— combining pose, emotion, and audio analysis for early Autism Spectrum Disorder (ASD) detection, and scaling LLM silver-label distillation for multilingual sentiment analysis. Before BHBFC, I led AI development at Peek AI (London), built VML's flagship AI product RevealWhich is Released EMEA region at Wunderman Thompson, and shipped AI systems for Japan Railway at BJIT Limited.
I am collaborating with researcher from HAAIM Lab,University of New Orleans,USA— On multi-modal based autism detection system using pose, emotion and audio analysis. I particulary helped them to provide the outline system design. Currently, i am focusing on Autonoumous Vehicle Research for South asia region. I am also open to freelance AI engagements, research collaborations,
Built VML's flagship AI product that analyzes campaign videos by detecting viewer emotions in real time. Combines emotion detection models, LLMs, Big Data pipelines, and Google Cloud Platform.
RAG-based AI platform helping doctors analyze medical reports, generate insights, and provide patient support with intelligent document understanding.
Multi-agent AI system automating end-to-end CRM operations using OpenClaw LLM and orchestrated agents for lead management, workflow automation, and customer follow-up.
AI book writer app that can write books on any topic including dark fantasy.It can generated images and can write books in any language. It is powered by uncensored AI model and can generate content without any restrictions. This software is built for group writer from USA who want automate their writing task.
RAG-based chatbot for the hotel industry — users book rooms, inquire about services, and manage reservations through natural language conversation.
RAG-powered chatbot for a government financial organization, assisting with loan inquiries, document guidance, and customer support in Bengali and English.
Desktop and web surveillance platform using object detection and pose estimation, optimized for CPU inference with OpenVINO for a London-based startup.
Real-time production throughput tracking for a matzoh bakery. YOLOv11 detection with ByteTrack persistence tracks dough balls through the class transition to matzah sheets, assigning per-worker stats via calibrated anchor zones on live CCTV feeds.
Intelligent worksheet generation platform for ESL teachers with AI-powered question and image generation, customized for language learning.
Complete smart hotel management — web platform + two mobile apps automating end-to-end operations from booking to in-room service with AI chatbots.
AI-powered virtual try-on system for fashion brands, enabling customers to visualize clothing using computer vision and generative models.
Developing the government's home-loan platform eSheba and associated mobile applications. Building FinTech modules with Java JEE, Spring Boot, Oracle Cloud, and Flutter.
Led development of AI CCTV surveillance using object detection and pose estimation. Built desktop and web apps with CPU-optimized inference via OpenVINO and YOLOv8.
Developed VML's MVP AI product Reveal — emotion-aware campaign analytics. Built ETL pipelines, dashboards, and backend services on GCP. Worked on micro-expression deep-learning models.
Developed a comprehensive ERP system for a Japan-based software company.
★ Top Performer (FY 2021-22)Built AI applications for Japan Railway — train detection, human activity recognition, and rail-crossing surveillance. Developed pose estimation and Bengali OCR using deep learning.
A scalable knowledge distillation framework that transfers GPT-4o-mini silver labels to XLM-RoBERTa for cost-effective multilingual sentiment analysis across 72.3 million social media comments.
Focuses on facial expression recognition for children with Autism Spectrum Disorder to enable downstream multimodal behavior analysis.
Served as subreviewer for the European Conference on Artificial Intelligence (ECAI) 2025, reviewing submissions in AI and machine learning.
I'm always open to discussing AI projects, freelance opportunities, research collaborations, or just a good technical conversation.
📧 md.sharif.ullah.forhad@gmail.com · 📍 Dhaka, Bangladesh