Ahmed Rafi Hasan

Ahmed Rafi Hasan (আহমদ রাফি হাসান in Bangla/বাংলা); Name of origin: Ahmed (Arabic) – Highly praised / Commendable | Rafi (Arabic) – Noble / Exalted | Hasan (Arabic) – Handsome / Virtuous

AI engineer & researcher

Striving to blend breadth and depth in AI.

Professional Email | Personal Email

I am an AI engineer and researcher specializing in Multimodal Machine Learning, Reinforcement Learning, and Computer Vision. Currently working as a Machine Learning Engineer at Pivotly, designing and developing Gen AI applications, LLMs, building MCP and other AI Agent protocols and end-to-end AI systems with cloud deployment expertise on Microsoft Azure.

My research focus spans Scene Graph Generation, Medical Image Analysis, and Precision Agriculture applications. I hold a B.Sc. in Computer Science and Engineering from United International University with Cum Laude distinction. I published research papers in IEEE Access, Computer and Electronics in Agriculture, IEEE TENCON, and other venues. Previously served as ML Engineer at Silcon Orchard Limited and Nodes Digital Limited working on Physics Guided Neural Net, Synthetic Data Generation, Enhanced Caption Generation, AgriTech solutions and IoT-based smart agriculture systems.

I am passionate about advance research in AGI development, with particular interest in interdisciplinary applications of AI. My long-term goal is to establish a world-class research lab contributing to safe and beneficial AI systems.

Industry Experience: LLM, Agentic Document Extraction, Retrieval-Augmented Generation (RAG), Multi-agent Systems (LangGraph, Autogen), Model Context Protocol (MCP) & Agentic AI Protocols, LLM Orchestration & Prompt Engineering, Natural-Language-to-SQL, Physics-Guided/Informed Neural Networks, Multimodal Image Clustering, Attention-based CNNs for Computer Vision, Model Deployment & API Integration (Azure).

Longterm Research Interest: Multimodal Machine Learning, Deep Reinforcement Learning, Embodied AI and Robotics(Multi-agent Control systems, Unmanned Aerial Vehicles), Physics Guided/Informed/Encoded Neural Network, 3D Scene Graph Generation, Multi-view synthesis and reconstruction, Medical Image processing, LLMs/VLLMs.

Currently looking for PhD opportunities. Open to talk about both academic and industrial collaborations and opportunities.

last updated: June 2026 (partial)

Updates

May 2026	Paper Accepted Paper accepted as poster at the 4th Workshop on Advances in Language and Vision Research (ALVR) of ACL 2026: Beyond Visual Similarity: Rule-Guided Multimodal Clustering with explicit domain rules.
May 2026	Paper Accepted Paper accepted as poster at the 4th Workshop on Advances in Language and Vision Research (ALVR) of ACL 2026: VLCE: An External Knowledge Framework for Contextual Image Captioning in Disaster Assessment.
Aug 2025	New Role Joined Pivotly as Machine Learning Engineer.
Mar 2025	Paper Accepted Paper accepted at Journal of Computers and Electronics in Agriculture[IF: 8.9]: Accurate water level monitoring in Alternate Wetting and Drying rice cultivation using attention-based ConvNeXt architecture.
Feb 2025	Promotion Promoted to Machine Learning Engineer at Silicon Orchard Limited.

Experiences

Professional Experiences

For details visit here.

Machine Learning Engineer at Pivotly
August 2025 - Present (Remote, Full-Time)

I’m currently involved in designing and developing AI-driven solutions across multiple areas:

Pivotly Parse
Developing agent-based document parsing architectures that can extract complex architectural drawings, floorplan layouts, and other structural informations and optimize workflow automation of business processes.
This systems includes custom prompt frameworks, fine-tuning segmentation and object detection models, Graph neural networks for optimizing floorplan layouts analysis and document parsing. This systems also includes multi-agent based architectures that is built on LangGraph and LangChain to provide agentic document parsing and analysis.

Multi-Agent Conversational Recommendation Systems (CRS)
Built a production multi-agent AI system (LangGraph + FastAPI) powering conversational product recommendation, agentic SQL query generation, and visual/semantic search for an e-commerce lighting catalog. Designed a manager agent that orchestrates specialized sub-agents (SQL agent, product agent, web-scraper agent) with structured outputs, dynamic state tracking, and contextual follow-up question generation to drive multi-turn recommendation flows. Implemented natural-language-to-SQL generation against a relational product database (PostgreSQL/MySQL/SQLite) with schema-aware prompting and async query execution, alongside image-embedding-based visual search and web-content summarization tools.

AI Search for Enterprise Use Cases
Developing intelligent search pipelines for client projects using natural language querying, vector search, and retrieval-augmented generation.

AI Feature Development for the Pivotly Platform
Creating and integrating AI components into Pivotly’s ecosystem, including workflow automation, building MCP and other AI Agent Protocols, LLM-based task execution, and intelligent UI interactions.

In-House AI R&D Projects
Working on internal prototypes around prompt frameworks, document intelligence, open-source LLM integration, and fine-tuning to build reusable AI modules.

Project Managers – Darren Gates, Byron Kwok, Kyle Karwatski

AriSaf Tech employee, Ahmed Rafi Hasan Junior Data Engineer and AI Researcher.

Machine Learning Engineer at Silicon Orchard Limited
February 2025 - July 2025 (Hybrid, Full-Time)

Worked on multiple in-house AI-based products, contributing to product development and client-focused AI solutions.

Developed and deployed a full Machine Learning pipeline for the product: SOLCHAT
This systems includes custom RAG pipelines with chatting interface which can update, delete vector chunks and create new vector chunks based on the user's request.
It is a scalabable production ready ML pipeline deployed wih Langchain, ChromaDB, FASTAPI and sqlite.
Worked on a Multimodal Sentiment Detection project
This systems includes tri modal(text, audio, video) fusion architecture for sentiment detection.
Contributed to VLLMs, LLMs, Explainable AI-based business solutions, and Class Incremental Learning
Developed multiple Azure-based ML services for clients

Junior Machine Learning Engineer at Silicon Orchard Limited
March 2024 - January 2025 (Full-Time)

Supervisors - Dr. Kishor Datta Gupta , Mohd Ariful Haque, Md. Saifur Rahman

Junior Machine Learning Engineer at Nodes Digital Limited
September 2023 - February 2024 (Onsite, Full-Time)

Developed ML algorithms for two funded precision agriculture projects in collaboration with Bangladesh Agricultural Research Council (BARC) and Institue for Advanced Research (IAR). Led the design and implementation of machine learning models for IoT-based smart irrigation systems, focusing on the Alternate Wetting and Drying (AWD) technique for sustainable rice production. Created predictive models for real-time water level monitoring and automated decision-making, enabling efficient water distribution among farmers. Worked on feasibility studies for IoT-based precision agriculture deployment across Bangladesh, contributing to both algorithmic development and system integration. Key achievements include developing computer vision models for water level detection, implementing edge computing solutions for IoT devices, and contributing to the development of a real-time water level monitoring system.

Machine Learning Intern at Nodes Digital Limited
January 2023 - August 2023 (Hybrid, Part-Time)

Supervisor - Dr. Swakkhar Shatabda

Ahmed Rafi Hasan research intern from CSIR-CDRI, Uttar Pradesh, India

Part-Time Lecturer/Instructor at Dept. of Computer Science & Engineering,United International University
June 2023 - September 2023 (Onsite, Part-Time)

Conducted and taught core undergraduate courses including Digital Logic Design and Discrete Mathematics, developing curriculum materials and conducting lectures for 100+ students. Designed assignments and mid/final qustions to assess and evaluated student performance, and provided academic mentorship.

Undergraduate Teaching Assistant at Dept. of Computer Science & Engineering,United International University
March 2022 - May 2023 (Onsite, Part-Time)

Selected as teaching assistant for multiple courses across four trimesters:
• Artificial Intelligence Laboratory - Guided students through ML/AI implementations and practical exercises for AI algorithms
• Data Structure & Algorithm I/II Laboratory - Conducted lab sessions on algorithm design and complexity analysis.
• Data Mining - Assisted with coursework on classification, clustering, and pattern recognition techniques and slide making for the course teacher.
• Structured Programming Language Lab - Taught C programming fundamentals and debugging techniques, designed final assignments and quizzes, and provided academic mentorship.

Responsibilities included conducting lab sessions, councelling students, grading assignments, holding office hours, and mentoring 200+ students in programming and algorithmic problem-solving.

Undergraduate Research Experience

Final Year Design Project on Improvement of Scene Understanding Using Scene Graphs
April 2022 - December 2022
Worked under the supervisions of Dr. Dewan Md. Farid and Akib Zaman
• Explored state-of-the-art scene graph generation techniques from recent research papers.
• Developed strategies for handling class imbalance in visual relationship detection datasets.
• Implemented and evaluated multiple architectures for object detection and relationship prediction and scene graph generation.

IAR Funded Research Project
November 2021 - December 2022
Worked under the supervision of Dr. Saddam Hossain Mukta
• Implemented the theoretical hypotheses of a Dynamic Weighted Tabular Method for CNNs.
• Conducted Image processing, training, and prepared ablation studies and result tables.

Volunteer Experiences

Student Volunteer Ahmed Rafi Hasan NeurIPS

Media & Communication Lead at UIU Robotics Club
March 2021 – May 2022
Coordinated effective communication with the departments & the DSA office, ensuring event planning and execution.Managed the club’s social media presence by creating and implementing innovative content strategies and engagement.

Collaorative Robotics Lab researcher Ahmed Rafi Hasan

Programming Wing Executive at UIU Computer Club
September 2021 – April 2022
Led counseling sessions during mid and final terms, offering academic support to students. Organized programming contests and conducted online workshops for freshers to build technical skills.

deeplearning.ai content tester, leader, volunteer Ahmed Rafi Hasan

Academic Member at Society for the Popularization of Science, Bangladesh (SPSB)
April 2019 – Dec 2019
Junior judge of Children Science Congress-2019(Science Projects) for evaluation of different types of scientific projects by young research enthusiasts. Instructor of Kudrat E Khuda Science Camp-2019. Conducted several science camps and workshops for Children Science Congress participants.

Publications

For details visit Google Scholar.

Published

Accurate water level monitoring in Alternate Wetting and Drying rice cultivation using attention-based ConvNeXt architecture
Ahmed Rafi Hasan, Niloy Kumar Kundu, Saad Hasan, Mohammad Rashedul Hoque, Swakkhar Shatabda
Published in - Computers and Electronics in Agriculture (Volume 234,Pages 110216)[Impact Factor: 8.9]

Accaptance Date: March, 2025 Read Here

A dynamic weighted tabular method for convolutional neural networks
Md Ifraham Iqbal, Md Saddam Hossain Mukta, Ahmed Rafi Hasan, Salekul Islam
Published in - IEEE Access (Volume 10,Pages 134183-134198)[Impact Factor: 3.6]

Accaptance Date: December, 2022 Read Here

A Novel Method for Imbalanced Data Classification based on Label Reassignment
Ahmed Shabab Noor, Afsana Airin, Rezab Ud Dawla, Ahmed Rafi Hasan, Muhib Al Hasan, Akib Zaman, Dewan Md Farid
Accepted in - IEEE Region 10 Conference (TENCON)(Page 1-6)

Accaptance Date: November, 2022 Read Here

Degrees of Anger Prediction from Speech
Ahmed Rafi Hasan, Akib Zaman, Puja Ghosh, Tazkia Tasnim Bahar Audry, Fardina Bhuiyan, Arindam Kundu Amit, Md Saddam Hossain Mukta
Accepted in - 25th International Conference on Computer and Information Technology (ICCIT)(Page 318-323)

Accaptance Date: December, 2022 Read Here

Attention-based scene graph generation: A review
Afsana Airin, Rezab Ud Dawla, Ahmed Shabab Noor, Muhib Al Hasan, Ahmed Rafi Hasan, Akib Zaman, Dewan Md Farid
Published in - 14th International Conference on Software, Knowledge, Information Management and Applications (SKIMA) (Page 210-215)

Accaptance Date: December, 2022 Read Here

As Co-Supervisor

Predicting Crop Prices using Machine Learning Algorithms for Sustainable Agriculture
Iftekhar Mahmud, Puja Rani Das, Md Habibur Rahman, Ahmed Rafi Hasan, Kamrul Islam Shahin, Dewan Md Farid
Published in - 2024 IEEE Region 10 Symposium (TENSYMP) (Page 1-6)

Accaptance Date: November, 2024 Read Here

Competitions

Best Performances

RSNA 2023 Abdominal Trauma Detection Top 25%. Team participation .
Participated in the medical imaging competition focused on detecting abdominal trauma from CT scans. Built a preprocessing and visualization pipeline, explored organ-level correlations, and implemented weighted log-loss based evaluation with scaling strategies to improve classification accuracy for multiple organ injuries.

UIU Intra University Deep Learning Sprint Champion. Team participation .
Developed a deep learning pipeline for validating image captions using pre-trained vision-language models (BLIP, ViT-GPT2). Implemented caption generation from given images and benchmark images, then evaluated semantic similarity with sentence transformers to produce validation scores. Designed and submitted the solution. .

AI Blitz⚡#8 by AIcrowd 9th position. Solo participation
Successfully completed all 5 F1-themed AI puzzles within the 3-week timeframe from car detection and speedometer reading to team classification, smoke removal, and car orientation classification.

ML Olympiad-CO2 Emissions Prediction Challenge 8th position. Team participation .
Applied different classical ML algorithms, time series and LSTM models to predict CO2 Emissions. .

Projects

Research Projects

Enhancing Loan Predictions
AI-based loan prediction system with enhanced support through loan policy and counseling services. The journey involved meticulous model training, integration of Retrieval-Augmented Generation (RAG) and model explainibility.

Taxi Car
A reinforcement learning project implementing Q-Learning to train a taxi agent in a 10x10 grid environment with obstacles and dynamic passenger/destination locations

Alzheimer Disease Detection from Brain MRI
Designed a system for Alzheimer’s disease detection using brain MRI, employing 3D medical image preprocessing. Trained 2D and 3D CNN models for accurate diagnosis. Applied image slicing, skull removal, Image registration and data augmentation.

Academic Projects

QucikChat
QucikChat is a dextop chatting app developed in Java. By using this application people can chat with friends in the network area(LAN) by using intranet mailing..

UIU Club Management System
Developed a university club management system to streamline membership, event scheduling, and activity tracking, enhancing administrative efficiency and student engagement.

Money Management App
It is a money management app where user can track his/her expenses in regular daily life. In our day-to-day life, it is important to track our expenses and savings and thus we would easily find out what things are needed further.