Building Systems. Advancing AI. Leading Teams.
MAHIR ASHAB
Applied AI Researcher, Backend Developer, and Campus Leader. Building data and AI systems for low-resource environments, edge intelligence, and scalable backend architectures — turning complex challenges into elegant solutions in code, research, and teams.
Engineer by Craft,
Leader by Instinct
I am a final-year Computer Science & Engineering student at Patuakhali Science & Technology University and an applied AI researcher building data and AI systems for low-resource, real-world settings. My work bridges edge intelligence for precision agriculture, multimodal deepfake detection, Bangla language models, and scalable backend services.
With a solid foundation in Data Structures & Algorithms, Object-Oriented Programming, and database systems, I approach every challenge with an architectural and problem-solving mindset. On the leadership front, I currently serve as President of both the PSTU Career Club and PSTU Entrepreneurship Club — driving university-wide career readiness, innovation programs, and industry engagement.
With three research papers submitted to ICCIT 2026 and an ICPC Regional ranking in the top 20%, I pair algorithmic rigor with hands-on systems deployment and cross-functional leadership.
Download ResumeICCIT '26 Papers
ICPC Regional Rank
Clubs as President
Screening Efficiency
Technical Arsenal &
Leadership Competencies
Technical Skills
Programming & Data
Machine Learning & AI
Models & Methods
Backend & Platforms
Systems, Edge & Tools
Leadership & Corporate
Languages
Core Strengths
Where I've Made
Impact
SmartAgro: Edge Intelligence for Crop Recommendation, Disease Detection & Adaptive Irrigation
AI & Embedded Systems Researcher
- ›Developed a fully local Raspberry Pi–Arduino edge AI system for crop recommendation, paddy disease detection, environmental sensing, and automated irrigation without dependence on continuous cloud connectivity
- ›Built a Random Forest + SHAP crop recommendation model across 22 crop classes and integrated sensor-driven inference into the local decision pipeline
- ›Fine-tuned MaxViT Tiny for 10-class paddy disease detection, achieving 97.60% single-pass test accuracy, and deployed image inference locally on Raspberry Pi for resource-constrained operation
FakeBDYouth: A Bilingual Deepfake Dataset of Bangladeshi Youth with Multimodal Detection
AI Researcher
- ›Developed a novel Bangla–English audiovisual deepfake dataset containing 3,345 clips from 30 participants across four authenticity classes
- ›Built a multimodal detection pipeline integrating ResNet50 for visual artifacts and Wav2Vec2 XLS-R for audio features, coupled with cross-modal attention and gated feature fusion
- ›Evaluated performance using 5-fold speaker-independent cross-validation, achieving 82.18% fusion accuracy, 88.20% manipulation-class F1, and 83.98% AUC
Evaluating Performance Recovery & Domain Robustness in INT8-Quantized BanglaBERT
AI Researcher
- ›Evaluated BanglaBERT model compression using unstructured pruning, dynamic INT8 quantization, cross-entropy recovery, and knowledge distillation
- ›Rigorously benchmarked predictive performance, model size reduction, CPU inference latency, serialization stability, and out-of-domain robustness
- ›Conducted empirical evaluation across 45,884 training samples, a held-out in-domain test set, and two challenging out-of-domain datasets
Academic
Foundation
B.Sc. in CSE
Patuakhali Science & Technology University
Relevant Coursework: Data Structures & Algorithms, Database Management Systems, Computer Networks, Operating Systems, Object-Oriented Programming, Software Engineering
HSC in Science
Notre Dame College, Dhaka
Milestones That
Matter
ICPC Regional 2024
Top 20%
Ranked 363rd out of 1,800+ teams in the prestigious International Collegiate Programming Contest regional round
ICPC Regional 2023
Competed across two consecutive regional appearances, demonstrating strong algorithmic problem-solving
Excellence in Management
Awarded for outstanding leadership and organizational contribution across university initiatives and student bodies
Verified in Real Time,
Not Just Claimed
Fetched live from Codeforces and GitHub public APIs. No manual updates — the data speaks for itself.
Codeforces
GitHub
What Drives Me
Beyond the Desk
Let's Build Something
Together
Whether it's a backend system that needs engineering, a team that needs leading, or an idea that needs executing — I'm ready to contribute.
Message Sent
Thank you for reaching out. I'll get back to you within 24 hours.
ashabmahir.200@gmail.com
Phone
+880 1768 363022
GitHub
github.com/Drogalion01
linkedin.com/in/ashab-mahir