Skip to main content

Md. Mehedi Hasan

Creating scalable AI solutions across visual intelligence, high-performance inference, and generative AI.

3.5+Years Experience
49Scholar Citations
2Open Source Contributions
  • Specialized in computer vision, inference systems, MLOps, and LLM applications.
  • Shipped AI agents, real-time vision pipelines, and optimized GPU inference workloads.
  • Bachelor in CSE, North South University
Resume
Picture of Mhasan502

Experience

AI Engineer

ProjsiteFull-time

Stockholm, Sweden

January 2025 - Present
  • Led development of Charlie, an AI support agent enabling real-time, context-aware interactions integrated with internal data and application state.
  • Designed and implemented agent pipelines (retrieval, tool use, response generation) to improve answer accuracy and reduce manual support dependency.
  • Contributed to the architecture and implementation of the next-generation Projsite platform, spanning backend, frontend, and AI systems.
  • Contributed to Magic Reader, enabling automated conversion of supplier order confirmations into actionable delivery plans using AI and OCR.
  • Built workflow orchestration using Apache Airflow, automating ingestion and transformation of external booking data into platform-ready formats.

Machine Learning Engineer

AlterSense LimitedFull-time

Dhaka, Bangladesh

Adviser: Dr. Nabeel Mohammed

January 2023 - December 2024
  • Architected and deployed a real-time ML pipeline processing ~1.1 GB/s of camera streams, integrating Apache Kafka for scalable distributed inference.
  • Delivered a robust object detection model in noisy environments, achieving an F1 score of 0.722 by addressing data imbalance challenges.
  • Optimized inference pipelines using TensorRT and NVIDIA Nsight, reducing GPU memory usage by 30% and increasing throughput by 1.7x.
  • Developed high-performance C++ ingestion modules using concurrency primitives (thread pools, mutexes, condition variables) for efficient streaming.
  • Built end-to-end data infrastructure (Airflow + data warehouse) to support automated ETL, real-time analytics, and scheduled batch inference.
  • Designed and scaled a CCTV-based data collection platform, reducing manual effort and improving data acquisition efficiency.
  • Developed a Temporal Tracking-based automated data collection platform, leveraging CCTV footage to reduce manual effort and optimize data collection.

Skills

Core

PythonC++PyTorchFastAPI

Data & Infra

MongoDBDockerRedisKafkaAirflow

Systems

Distributed PipelinesConcurrencyETL OrchestrationEvent-Driven Systems

Specialization

AI AgentsLLM AppsReal-Time Vision InferenceVision Optimization

Education

Bachelor of Science in Computer Science and Engineering

Spring 2018 - Summer 2022

Department of Electrical and Computer Engineering, North South University

  • CGPA: 3.77 / 4.00 (90-92% marks)
  • Graduated with Magna Cum Laude.
  • Thesis Title: Inter-Dataset Critical Evaluation of Common Object Detection Model.
  • Achieved 2nd Runner Up position in Electrathon 2018 organized by IEEE NSU.

Publications

1Results in Engineering 18 (2023): 101079

“LULC changes to riverine flooding: A case study on the Jamuna River, Bangladesh using the multilayer perceptron model”

Md Mehedi Hasan*, Md Sahjalal Mondol Nilay, Nahid Hossain Jibon, Rashedur M. Rahman

  • Land-Use Land-Cover (LULC) was generated by supervised classification from Landsat 5 and Landsat 8 data.
  • The land-use land-cover changes were aggregated in a Markov matrix for land-use land-cover change analysis.
  • The Riverine Flood Potential Index (RFPI) was calculated using the Multilayer Perceptron (MLP) model for both 1990 and 2020.
  • LULC changes and riverine flood potential changes were calculated using the total relative difference formula.
  • Geographically Weighted Regression (GWR) derived a statistical correlation between LULC and riverine flood potential changes.
Results in Engineering 18 (2023): 101079