I am a Master of Data Science (Professional) student at Deakin University with a background in Computer Science and Engineering and practical experience across Artificial Intelligence, Machine Learning, Deep Learning, Data Analytics, Data Engineering, and MLOps.
My work focuses on building intelligent and production-oriented systems, including multi-agent AI, RAG pipelines, LLM applications, predictive modelling, computer vision, time-series forecasting, and cloud-deployed ML systems.
I have professional AI engineering experience building agentic systems with LangChain, LangGraph, Django REST API, RAG, and Azure AI Studio, alongside academic research in few-shot NLP, SETI signal classification, computer vision, and machine learning.
I am particularly interested in:
- Artificial Intelligence & Machine Learning
- Deep Learning & Neural Networks
- Generative AI & LLM Applications
- Multi-Agent AI Systems
- Retrieval-Augmented Generation (RAG)
- MLOps & ML Engineering
- Computer Vision
- NLP & Question Answering
- Time-Series Forecasting
- Data Analytics & Statistical Modelling
- AI applications in Space & Scientific Computing
Deakin University — Master of Data Science (Professional) Melbourne, Australia | Feb 2026 – Nov 2027
- Deakin International Meritorious Scholarship — 25%
North South University — B.Sc. in Computer Science and Engineering Dhaka, Bangladesh | Jan 2019 – Apr 2025
Python R Java SQL C C++ Fortran
Machine Learning Deep Learning NLP LLMs RAG
Multi-Agent AI LangChain LangGraph
TensorFlow PyTorch Scikit-learn Hugging Face Transformers
Statistical Analysis Predictive Modelling Time-Series Forecasting
Feature Engineering A/B Testing Data Analysis
Azure AWS Docker Kubernetes Azure ML
Vertex AI AWS SageMaker AWS Lambda
Azure AI Studio OpenAI API GitHub Actions CI/CD
PostgreSQL MongoDB MySQL Snowflake Databricks
Django FastAPI React.js Node.js
Sep 2024 – Mar 2025
- Built a multi-agent AI system for B2B supply-chain operations using LangChain and LangGraph.
- Deployed AI workflows through Azure AI Studio, reducing vendor-matching time by approximately 35%.
- Integrated real-time application updates through Django REST API.
- Improved retrieval accuracy by approximately 18% through RAG and prompt tuning.
- Designed agent handoffs between planning, execution, and verification components for production-oriented agentic workflows.
Multi-agent policy-document review system developed as part of Google's Agents Intensive Capstone.
Technologies:
Google ADK Gemini 2.5 MCP FastAPI Docker Google Cloud Run
- Architected three collaborating agents: Monitor, Authoriser, and Comparison.
- Integrated Google Drive and Gmail tools using OAuth and service-account authentication.
- Built robust error-handling workflows.
- Containerised and deployed the application using FastAPI/Uvicorn and Google Cloud Run.
Time-series forecasting project focused on NVIDIA stock-price prediction.
Technologies:
Python LSTM XGBoost ARIMA Time-Series Analysis
- Collected market data using Alpha Vantage.
- Performed preprocessing, stationarity testing, seasonal decomposition, and feature engineering.
- Compared LSTM, XGBoost, and ARIMA forecasting approaches.
- LSTM achieved the best reported performance with a MAPE of 1.32%.
Research project focused on improving transformer-based question answering in low-resource settings.
Technologies:
Python PyTorch Transformers BERT RoBERTa MoE BPE
- Investigated Dynamic Multi-Head Attention, simplified Mixture-of-Experts, and Learnable Positional Encoding.
- Developed a 10k-vocabulary BPE tokenizer and span-alignment pipeline.
- Experimented with BERT, RoBERTa, DistilBERT, ALBERT, SpanBERT, and Splinter.
- Applied FP16 mixed precision, gradient clipping, dropout, cosine annealing, and early stopping.
Deep-learning research project for classification of radio-signal spectrograms.
Technologies:
Python TensorFlow/PyTorch CNN InceptionResNetV2 ResNet50 MobileNetV2
- Developed a three-branch CNN ensemble for SETI radio-signal classification.
- Used STFT spectrograms, SpecAugment, class balancing, and multi-seed evaluation.
- Implemented Grad-CAM for model interpretability.
- Classified 7 radio-signal categories across 7,000 labelled train/validation/test samples.
- Achieved approximately 95% classification accuracy.
Detection of Black Holes from Stellar Datasets Using Machine Learning
H. R. Fahim, M. Hasan, K. M. A. Salam. 2024 IEEE International Conference on Future Machine Learning and Data Science (FMLDS), pp. 468–473, IEEE, 2024.
Developed a dual-sided campus food-sharing marketplace.
Technologies:
React 19 TypeScript Vite Tailwind Firebase Firestore
- Implemented real-time Firestore synchronisation.
- Developed Customer and Cooker role-based interfaces.
- Implemented Google OAuth university-email verification.
- Built ordering, pricing, trust-and-safety, strike/ban, and live-feed functionality.
Completed entrepreneurship and startup-development training covering:
- AI-driven innovation
- Customer discovery
- Business validation
- Prototyping
- Pitch development
- Lean startup methodology
- Venture creation
🔹 Artificial Intelligence — Generative AI, Agentic AI & Business Automation
🔹 Machine Learning — Predictive Modelling & Scientific Applications
🔹 Deep Learning — Neural Networks & Representation Learning
🔹 Computer Vision — Medical Imaging, Robotics & AR/VR
🔹 NLP — LLMs, RAG & Question Answering
🔹 Space AI — SETI, Radio Signals & Astronomical Data
🔹 MLOps — Production ML Systems & Cloud Deployment
- SEED-AI — Skills for Education, Employment and Digital AI, Deakin SME Research Centre, 2026
- Neural Networks and Deep Learning, Coursera, 2025
- Advanced Learning Algorithms, Coursera, 2025
📍 Melbourne, Victoria, Australia