Hi, I am Abrar Eyasir. I am currently in the final year of my undergraduate studies in CSE at the University of Dhaka, Bangladesh. In my first year, I co-founded Sirius Academy, where I have taught and mentored over 500+ students. I always try to maintain a balance between my academic work (CGPA: 3.84/4.00) and extracurricular activities.
In my third year, I developed a strong interest in hackathons. I have actively participated in several competitions with my team, DU_Caffeine. We won the Solvio AI Hackathon 2025 with our project ZeroBin: AI-Driven Waste Management Platform and were also placed among the top 10 teams in a number of national hackathons, including InnovativeX and VisionX.
Over time, my interest has gradually shifted towards research. My undergraduate thesis focused on developing a large-scale Bangla QA dataset and fine-tuning transformer-based models for question answering, under the supervision of Dr. Muhammad Ibrahim. Currently, I am exploring areas such as Natural Language Processing (NLP), Large Language Models (LLMs) and generative models like Stable Diffusion.
I also enjoy writing articles, creating video content (3300+ Youtube Subscribers) and sharing what I learn with others. Recently, I was selected as a Section Leader for Code in Place 2026, where I will be teaching Python to students around the world. I have also been selected for the NUS Young Fellowship Programme 2026 and will be attending the programme at NUS this summer
In the long run, I hope to grow into a research-driven entrepreneur in AI/ML/DL and contribute to meaningful work in this field. I know I still have a lot to learn and I am working on improving myself every day.
🔥 News
- 2026.04 Selected as a Section Leader at Code in Place 2026 by Stanford University
- 2026.03 Selected at NUS Young Fellowship Programme 2026
- 2025.12 Champion at Solvio AI Hackathon with Project ZeroBin: AI-Driven Waste Management Platform
- 2025.11 Top 20 at VisionX Project Showcasing 2025
- 2025.09 7th Place at InnovateX Hackathon, BUBT
- 2025.07 Completed Summer School on Energy Data Analytics at IIT Bombay, India
- 2025.03 Started working as a Research Assistant at the University of Dhaka under Dr. Muhammad Ibrahim
📝 Research Papers
💼 Experience
- Developed a large-scale Bangla QA dataset with 94,000 QA pairs and 210M tokens, implementing an automated pipeline for document extraction, segmentation, and QA generation
- Fine-tuned Qwen-2.5-7B using PyTorch, PEFT LoRA, and Chain-of-Thought prompting for parameter-efficient large model training
- Implemented fine training for transformer-based extractive QA models including BERT, RoBERTa, and ELECTRA
- Benchmarked generative vs. extractive QA systems using BLEU, F1, and Exact Match metrics on large-scale datasets
- Core ML team member, developing and deploying ML solutions across 7+ national hackathons
- Built ML systems including computer vision, NLP models, and time-series forecasting pipelines
- Engineered scalable deployments using FastAPI, Docker, and cloud infrastructure
- Awards: Champion – Solvio AI Hackathon; Top 20 – VisionX; 7th Place - InnovativeX
🎓 Education
🎓 Summer Programme
Completed intensive coursework on large-scale energy-sector datasets. Applied machine learning and time-series analysis to extract actionable insights from real-world scientific data (ASHRAE Building Energy Dataset).
🚀 Selected Projects
Sep-Dec 2025
ZeroBin — AI-Driven Waste Management Platform
Machine Learning Engineer & Team Lead
Live Demo GitHub 🤗 Hugging Face
- Led 4-member team to design and deploy ML-driven solutions for smart city waste management
- Developed LightGBM geospatial prediction and LSTM time-series forecasting models for resource optimization
- Fine-tuned Onnubuti AI sentiment analysis model achieving 94.5% accuracy
- Deployed scalable inference system with FastAPI, Docker, and RESTful APIs
- 🏆 Champion at Solvio AI Hackathon (2025)
Dec 2025-Jan 2026
Distributed Machine Learning System with MPI & OpenMP
- Developed a distributed ML system using C, MPI, and OpenMP with synchronized gradient updates across 8 parallel processes
- Built a low-latency inference pipeline integrating MPI with FastAPI backend and Next.js frontend for real-time predictions
- Implemented parallelized matrix multiplication with OpenMP to enable multi-linear regression
Oct 2024-Jan 2025
Flash — Tutorial Recommendation Platform
- Implemented content-based recommendation system using TF-IDF and cosine similarity
- Built full-stack application for personalized tutorial recommendations
🎖 Awards & Achievements
💻 Technical Skills
📫 Contact
I am actively seeking opportunities for graduate studies and research positions in Machine learning, Large Language Models(LLMs) and Natural Language Processing(NLP).
Feel free to reach out — I'd love to connect!