AI Agent & IoT Researcher · Available for PhD / MSc 2026–27
Md. Hasnain Ali
AI Agent & IoT Researcher · Autonomous Systems · Edge Intelligence

I design and build intelligent agent systems for IoT ecosystems — combining reinforcement learning, computer vision, and edge AI to create autonomous, real-time decision-making systems. My work bridges embedded systems and machine learning, with a focus on deploying intelligent agents in resource-constrained environments.

Download CVView ResearchGet in Touch
2
Publications
8+
Open-Source Projects
3+
Years of Research
2
Years of Industry
01 — Research
Research Interests
AI Agents & Autonomous Systems
Designing multi-agent systems for autonomous decision-making, task planning, and tool-use. Research on agentic workflows, LLM-based reasoning, and self-improving agent architectures for real-world deployment.
IoT & Edge Intelligence
Intelligent IoT systems with on-device AI inference, sensor fusion, and real-time data processing at the edge. Focus on low-power, latency-critical deployments for smart environments and industrial IoT.
Computer Vision for IoT
Real-time object detection, pose estimation, and video analytics optimized for edge devices. YOLO architectures, lightweight vision transformers, and privacy-preserving on-device vision pipelines.
Reinforcement Learning & Robotics
Deep RL for robotic control, sim-to-real transfer, and autonomous navigation. Integrating perception and action loops for embodied intelligence in dynamic environments.
Medical AI & Healthcare IoT
Deep learning for automated medical diagnostics, retinal image analysis, and explainable AI. Building IoT-enabled healthcare monitoring systems with edge-based inference.
MLOps & AI Infrastructure
Scalable ML pipelines, model versioning, containerized deployments (Docker/K8s), CI/CD for AI systems, and efficient inference serving for production-grade agent platforms.
02 — Publications
Research Output
[P1] · Journal
Interpreting Deep Neural Networks in Diabetic Retinopathy Grading: A Comparison with Human Decision Criteria
S. Biswas, M. A. A. Khan, M. H. Ali, J. Rohdin, S. Pramanik, M. I. A. Khan, S. K. Chakravarty, B. K. Pramanik
Life, Vol. 15, No. 9, Article 1473 — September 2025
Contribution: Co-author; contributed to model interpretation experiments comparing deep neural network decisions with human grading criteria for diabetic retinopathy.
A study interpreting deep neural networks used for diabetic retinopathy grading by comparing their decision criteria with those of human clinicians, providing insight into how DNN predictions align (and diverge) from expert ophthalmologist assessments.
↗ Paper
Medical AIDiabetic RetinopathyExplainable AIDeep Learning
[P2] · Conference
Competency Comparison of Deep Neural Networks for Identifying Gender in Color Fundus Photographs
M. A. A. Khan, M. H. Ali, N. Saha, M. S. S. Shoumik, S. Biswas
26th International Conference on Computer and Information Technology (ICCIT), Dhaka, Bangladesh — 2023
Contribution: Co-author; contributed to experimental design, model training, and evaluation of deep neural networks for gender identification from color fundus photographs.
A comparative study evaluating the competency of multiple deep neural network architectures in identifying gender from color fundus photographs, an important preprocessing step in ophthalmic AI systems that can introduce bias if not handled carefully.
↗ Paper
Medical AIOphthalmologyFundus ImagingDeep Learning
03 — Engineering
Projects
AI Agent Interview Platform
Apr 2025
Built backend and AI infrastructure for an end-to-end AI-driven interview platform with real-time audio-based assessment. Reduced per-interview AI cost by 90% through prompt caching and context compression. Integrated Whisper Large v3 STT and open-source TTS.
FastAPIRedisWhisperLLM AgentsPostgreSQL
⌥ GitHub
VisionDesk: AI Workforce Monitoring
Mar 2025
AI-driven workforce activity monitoring with real-time pose estimation and activity recognition. Fine-tuned models for productive vs. non-productive behavior classification. Django REST API with Redis-backed async task queue for video stream processing.
DjangoFastAPIYOLOMediaPipeComputer Vision
⌥ GitHub
Real-Time Surveillance for Safety
Nov 2024
Multi-object detection system trained on 13K+ images for construction worker safety. Benchmarked 8 YOLO architectures (v8-v11), achieving 92% [email protected]. Optimized inference to 15 FPS on 640x640 streams for real-time edge deployment.
YOLOv8YOLOv11PyTorchEdge AIComputer Vision
⌥ GitHub
AI Retinal Blood Vessel Segmentation
2024
Custom U-Net model for accurate retinal blood vessel segmentation in medical IoT diagnostics. Streamlit-based interface for real-time inference, deployed as an edge-capable medical AI tool.
U-NetTensorFlowMedical AIIoTEdge Inference
⌥ GitHub
Image Similarity Detection Engine
Feb 2025
Sub-millisecond image similarity pipeline using ResNet-50 and FAISS with GPU acceleration on 100K+ images. Achieved 70% computational reduction via PCA. Built for large-scale visual search in IoT camera networks.
PyTorchResNetFAISSPCACUDA
⌥ GitHub
BanglaShift: AI Translation Agent
Jan 2025
Fine-tuned Llama 3.1 8B for Banglish-to-Bangla translation achieving 95% accuracy. FastAPI microservice with MLflow tracking, 200ms response time. Demonstrates LLM agent deployment for language tasks.
Llama 3.1FastAPIMLflowLLMDocker
⌥ GitHub
04 — Experience
Experience
Software Engineer L-I (AI/ML)
Vivasoft Limited — Dhaka, Bangladesh
Apr 2025 – Present
  • Developing AI agent modules for interview automation, integrating Speech-to-Text (Whisper) and Text-to-Speech pipelines supporting 1,000+ monthly evaluations.
  • Engineering prompt optimization strategies for LLM-based agents, improving response relevance and contextual understanding in production.
  • Building scalable AI infrastructure with FastAPI microservices and Redis-backed task queues for real-time agent interactions.
Associate Software Engineer (AI/ML)
Vivasoft Limited — Dhaka, Bangladesh
Sep 2024 – Mar 2025
  • Led R&D initiative replacing GPT-based API calls with fine-tuned in-house LLM for structured data extraction, achieving 90% cost reduction.
  • Curated and annotated custom training dataset of 5,000+ examples; fine-tuned open-source LLMs for production-grade agent tasks.
  • Researched and deployed efficient inference strategies for resource-constrained environments.
Student Research Assistant
BioMe Lab — University of Rajshahi
Jan 2023 – Oct 2025
  • Conducted research in medical AI, computer vision, and IoT-enabled diagnostics under faculty supervision.
  • Developed deep learning models for retinal image analysis and automated medical diagnostics.
  • Supervisor: Sangeeta Biswas, Associate Professor, University of Rajshahi.
05 — Achievements
Achievements
Champion
AI Hackathon 2025
1st place among competing teams at the AI Hackathon organized by Bangladesh Innovation Conclave & Brac University — beating field with end-to-end AI agent build under time pressure.
May 2025 · Bangladesh Innovation Conclave × Brac University
Finalist · Top 7
Datathon 3.0
Selected among the top 7 finalists at Datathon 3.0 organized by Robi Axiata Limited — a national-scale data science competition.
June 2024 · Robi Axiata Limited
1st Runner-Up
Nacter Robotics Olympiad — Poster Presentation
1st Runner-Up at the poster presentation track of the Nacter Robotics Olympiad, organized by the National Academy for Computer Training & Research.
August 2022 · NACTER
Finalist
BitFest AI Hackathon
Finalist at the BitFest AI Hackathon organized by the Department of CSE, KUET & Brain Station 23.
January 2025 · KUET × Brain Station 23
06 — Background
About

I'm an AI/ML engineer based in Dhaka, Bangladesh with a B.Sc. in Computer Science and Engineering (CGPA 3.56/4.0) from the University of Rajshahi, where I spent three years as a research assistant in the BioMe medical-imaging lab.

Today I work on the production side: fine-tuning open-source LLMs to replace commercial APIs, shipping real-time computer vision systems that run on constrained hardware, and building AI agents that handle real workloads — like an interview platform serving 1,000+ candidates a month.

I'm looking for a PhD position (Fall 2026/27) where I can push this work into the research frontier — agentic systems, efficient inference, and AI that holds up outside the lab.

Bachelor of Science — Computer Science & Engineering
University of Rajshahi
2021 – 2025 · Focus: AI, Machine Learning & Embedded Systems
Higher Secondary Certificate (HSC)
Notre Dame College, Dhaka
2019 · Science Group
07 — Technical Background
Skills & Tools
AI Agents & LLMs
Llama / Mistral Fine-TuningLangChainRAG PipelinesPrompt EngineeringUnslothWhisper STT / Piper TTS
Computer Vision
YOLOv8–v11OpenCVObject DetectionImage SegmentationVision TransformersPose Estimation
Deep Learning
PyTorchTensorFlowHuggingFaceScikit-LearnXGBoostOptuna · Ray Tune
Backend & APIs
FastAPIDjango RESTFlaskRedisPostgreSQLDocker
MLOps & Cloud
MLflowAWS SageMakerAzure MLGCP AIGitHub Actions
Languages
PythonC++JavaSQLBashGit
08 — Writing
Blog
Mar 2025 · ◷ 12 min
Building Autonomous AI Agents: A Practical Guide
A hands-on walkthrough of designing LLM-powered agents with tool-use, memory, and self-reflection loops. Covers architectures, prompting strategies, and production deployment patterns.
AI AgentsLLMsArchitecture
Feb 2025 · ◷ 8 min
Real-Time Video Analytics on Edge Devices
How to deploy YOLO-based object detection on low-power edge devices for real-time surveillance. Optimization techniques for achieving 15+ FPS on ARM-based hardware.
Edge AIYOLOIoTOptimization
Jan 2025 · ◷ 8 min
From Engineering to Research: My PhD Journey
Honest reflections on transitioning from AI engineering to academic research in AI Agents and IoT — what skills transfer, what gaps exist, and how to write a competitive statement of purpose.
ResearchPhDCareerAI Agents
Dec 2024 · ◷ 10 min
IoT-Enabled Medical Diagnostics with Edge AI
Building end-to-end healthcare IoT systems with on-device deep learning inference. Case study on retinal image analysis deployed at the edge for real-time diagnostic support.
IoTMedical AIEdge ComputingDeep Learning
09 — Contact
Get in Touch
Location
Dhaka, Bangladesh
Find Me On
⌥ GitHub↗ Google Scholar↗ LinkedIn