Mijanur Palash, Phd Email & Phone Number

West Lafayette, In, Us

Oak Ridge, Tennessee, United States

Train LLMs and vision transformers (ViT) on word's largest and only exa-scale supercomputer.

Looking for Job in: Machine Learning & Deep Learning, Software Engineering

West Lafayette

1)Multimodal Emotion Detection:-Developed (in Python) emotion detection system from video and image data (modes: face, posture, and gait) using deep learning techniques- CNN and 3DCNN, achieved high accuracy (75%) in FER-2013 dataset-Analyzed bias in existing emotion datasets and proposed a new dataset collected with web scrapping and crowdsourcing 2)Deep Learning Based Autism Detection-Designed (in Python) a system for Autism detection from multimodal video/image data. It achieved 92% accuracy using Ensemble Convolutional Neural Network (ECNN) and transfer learning with pre-trained RESNET-503)Machine Learning for Situational Knowledge on Demand-Designed (in Python) video object detection and feature extraction module using OpenCV and CNN. Created rooted localized knowledge graph, used PostgreSQL database to store the features (NGC funded)-Performed textual feature extraction from police incident report using Word2Vec and BERT techniques 4)Reinforcement Learning Based Collusion Detection-Trained (in Python) RL agent to play, collude and detect other players’ collusion in Leduc Poker game using Soft Actor-Critic (SAC) and Deep Q Network (DQN) techniques (DARPA funded)Robust 360 Video Streaming -Designed ‘CoRE’- a system (in C++) for 360◦ video streaming, achieved 30% bandwidth reduction (NSF funded)ML Projects: 1) Implement and compare Naive Bayes, Logistic Regression, SVM, Decision Trees, Bagged Trees and Random Forest algorithms on Speed Dating dataset 2) Implement Support Vector Machine (SVM) on Iris dataset 3) Implement CNN algorithm on MNIST dataset 4) Implement GNN with mean and LSTM aggregator 5) Implement Markov Chain and LSTM language modelling 6) Implement and compare different VAE architectures 7) Implement a GAN for image generation and discrimanation

Santa Clara, California, United States

Interactive Nerf Camera (INF):• Developed a front-end interface for facilitating the generation of interactive NeRF datasets and enabling interactive playback• Enabled user control of viewpoints and exploration of different perspectives within the NeRF framework• Developed the back-end system for automatic remote training and inference, enabling seamless training of NeRF modelsNeural Frame Extrapolation (NFE): • Improved the preprocessing script for NFE to generate required warp buffers for training• Implemented multiprocessing to increase the batch size by 8x, reducing training time• Employed automated mixed precision (AMP) to optimize GPU memory usage during training• Regularly generated training data, performed training and inference for parameter tuning and optimization, ensuring continuous improvement of the NFE module

Carbondale Illinois

CGPA: 4.0/4.0Achievements:• Conducted research on the effect of Multipath TCP in WiFi and LTE network’s throughput and energy aspects.• Proposed MPWiFi scheme for better Wi-Fi performance and implemented in Linux kernel.• Developed KMPTCP system, which can automatically switch between TCP and MPTCP to save energy and improve throughput. Used C++, UNIX programming and Shell scripting to develop KMPTCP. • Built and maintained networking testbed in the SIUC Networking Lab. Gained experience in SDN and Open-flow, NFV, routers switching and forwarding, VLAN switches, security and privacy concepts including deep packet inspection, onion routing, data encryption, IDS an firewall etc. • Excelled in modeling and simulating networks using simulation software Riverbed, NS3 and Mininet and troubleshooting using packet analyzer Wireshark and Tcpdump. • Analyzed Lemko private LTE base station’s performance for mission critical machine-to-machine and IoT applications and proposed solutions to improve fluctuation in round-trip-time (RTT) using MPTCP.• Helped PI to get 30,000 USD research grant from LEMKO Corporation USA.

Carbondale

- Conducted lab session for ECE 222 (C/C++) course. Helped sophomore students to write and debug code in C and C++. - Conducted lab session for ECE 296 (Software Tools) course. Helped students to write and debug code in Python and Matlab and to work with Raspberry Pi and Arduino microcontroller. Worked closely with students for their final projects to build a line follower robot capable of traveling through a maze.

Doctor Of Philosophy - Phd, Computer Science

Master’S Degree, Computer Systems Networking And Telecommunications

1. Novel Wi-Fi Networking with Multipath TCP 2. Private LTE network for M2M communications

Bachelor Of Science (Bsc), Electrical And Electronic Engineering

Activities and Societies: BUET Literature Club, BUET Debating Club and Badhon- a voluntary blood donating organisation.

Higher Secondary Certificate Exam, Science

Activities and Societies: Notre Dame Debating Club, Notre Dame Nature Study Club