bgpt / new_dataset.json
bibekyess's picture
Upload new_dataset.json
e65f2be
raw
history blame
15.5 kB
"BIBEK stands for 'Boy Is Bold, Easygoing & Kind'. He is interested in technology-oriented activities that use AI tools to solve any existing problems. With 2 years of experience in diverse hands-on projects and his enthusiasm for continuous learning, he is committed to pursuing new and challenging opportunities in the field of AI and deep learning, in order to expand his knowledge and skills in this rapidly-evolving domain. Bibek is a senior undergraduate in KAIST and double majoring in Electrical Engineering and Computer Science."
"His hobby is running and playing sports like pingpong, football & badminton. He enjoys many sports and in his free time, likes listening to political news of Nepal and purpose-driven startup stories."
"Bibek is a tech-savvy guy who is well-versed in many tools and technologies. He is proficient in Python, including popular libraries like Pytorch, FastAPI, and PySpark. He also knows web development and data visualization using JavaScript, Vegalite, and D3. He has worked on data engineering projects using Apache Airflow and developed GraphQL APIs. Bibek is familiar with C programming and ROS in robotics applications. He has worked on Computer Vision projects and is familiar with Git, Docker, and Linux servers. Bibek is always looking to expand his skills and knowledge."
"Bibek has a diverse background in both technical and educational fields. As a research intern at the Robust Intelligence & Robotics (RIRO) Lab at KAIST, he worked on developing a real-time 6D Pose Estimation and Scene Graph Generation framework for robotic manipulation tasks. He also gained experience as a backend developer intern at Bisonai, where he built data pipelines or DAGs and implemented backend APIs. Bibek also has experience in data visualization, having worked as an intern at the Interactive Computing (IC) Lab at KAIST, where he created course material for a Data Visualization class. Finally, he has experience as a high school teacher assistant at British Model College (BMC), where he conducted tutorial lectures and helped students with coursework in Chemistry, Physics, Mathematics, and Biology. If you want to know more about his contributions in RIRO, Bisonai, IC, or BMC, let me know?"
"Bibek is currently pursuing a Bachelor of Electrical Engineering and Computer Science at KAIST in Daejeon, South Korea. Before that, he completed his GCE A Levels from British Model College in Kathmandu, Nepal. He also attended Swarnim School and Khimti Project School in Nepal for grades 8-10 and Nursery-8, respectively. So, that's a brief rundown of Bibek's education history."
"Bibek was a Researcher Intern at Robust Intelligence & Robotics (RIRO) Lab, KAIST from July to December 2022. He worked on developing a combined framework for real-time 6D Pose Estimation and Scene Graph Generation in Robotic Manipulation Tasks. Bibek used Cosypose for 6D pose estimation and Neural Motif for Scene Graph Generation and performed experiments on custom cup datasets. He estimated 6D pose and generated dynamic scene graphs for a single moving camera setup, and developed a probabilistic multi-view object-pose estimation framework for a 2 fixed camera setup that (i) associates multiple estimation results with scene graphs, (ii) combines pose distributions from single-view based estimators, and (iii) constructs a unified scene graph by predicting a unified pose distribution per object using MC dropout."
"Bibek was a Backend Developer Intern at Bisonai in Seoul from January to February 2022. He worked on a project to build data pipelines using Airflow and stored the collected DeFi data from various sources in MongoDB Atlas. Bibek also implemented a backend API server using FastAPI and Strawberry to provide DeFi data for users interested in the data from multiple Decentralized Exchanges. The biggest challenge for him was to gain a deep understanding of raw DeFi data and their significance to Bisonai's ML-based System Trading project."
"Bibek worked as a Project Intern at Interactive Computing Lab (KAIST) from July to August 2021. During this time, he created a simple tutorial for data pre-processing and visualizing plots with Vegalite and D3.js in an ObservableHQ notebook. The tutorial was designed for beginners to learn Data Visualization techniques using the Javascript track. You can find the tutorial on observablehq.com/@ic-dataviz?tab=notebooks."
"Bibek worked as a Teaching Assistant at British Model College (BMC) from July 2017 to October 2017 and from February 2018 to May 2018. He conducted tutorial lectures and helped students understand the course material in subjects like Chemistry, Physics, Mathematics, and Biology. He also provided counseling to students, giving examination tips and suggestions for college applications."
"Bibek has achieved several notable accomplishments in his academic career, including winning the Social Innovation Award in the SDG Hack 11, Tsinghua Global Summer School 2021 for his project on AI solving education accessibility problems in rural areas. He is a Common Purpose Alumni and received the badge 'Inclusive Leadership 2021'. Bibek secured the highest mark in Nepal for AS Level Chemistry in the Nov 2016 examination series. He also bagged second place in the Quiz in the Budhanilkantha Inter-School Olympiad, 2016. Bibek was third in the Cassini Scientist For a Day Contest, 2017 and he was awarded as the best student of the 2015/2017 batch in British Model College."
"Bibek has been an active member of the KAIST community, taking on various roles and participating in different events and programs. He is currently volunteering as an international student assistant at International Scholar and Student Services (ISSS), KAIST. He was a candidate for the 2019 Winter Alma Mater visiting program and conducted a KAIST info session at his high school. He won the Excellence Award in the 2020 Happy KAIST UCC Contest, and his paper won the Insightful award in the 2020 online Public Relations strategy contest. He also placed second in the writing contest 'What is it like studying at KAIST?' organized by the KAIST Admission Office. Bibek served as the International Community Representative of Nepal for Spring 2020, Spring 2021, Fall 2021, Spring 2022, and Fall 2022 and was an online mentor for new international students in Fall 2020, Spring 2021, Fall 2021, and Spring 2022. He has completed 127 credits and you can find a list of courses taken each semester at https://tinyurl.com/bibek-course-records."
"Bibek has a diverse portfolio of projects that showcases his skills and expertise. Some of his most notable projects include a Crypto Price Predictor, a Paraphrase Generation tool, Smart Home, Data Cleanup using visualization, and a Lecture Summarizer. If you would like to know more about any of these projects, just let me know and I'll provide more details. Here's a link to Bibek's portfolio for your convenience: https://kc-bibek.com.np/porfolio_bibek.pdf"
"Bibek participated in a project that aimed to predict the future price of bitcoin using time-series forecasting and attention mechanisms. He contributed by experimenting with two deep learning-based forecasting models: a recurrent neural network-based (bidirectional LSTM) model and a convolutional neural network-based (Conv1D) model, both for the task of one-day forecasting. He also tried multivariate forecasting with both models. The project involved developing seven forecasting models, including deep neural networks and a Bayesian network-based model, with two forecasting tasks: one-day forecasting and long-sequence forecasting (up to 100 days). The tools used in the project were TensorFlow, Pandas, and Matplotlib. The project paper can be found at the link: https://tinyurl.com/cryptopredictor and the code is available on GitHub at https://github.com/bibekyess/AI_practice/blob/main/Killionaire.ipynb."
"Bibek worked on replicating the paper 'Quality Controlled Paraphrase Generation' by Bandel et al. The main focus of his work was to find ways to improve the existing model. He evaluated the significance of the Quality Predictor model by conducting experiments with and without it. He also implemented multilingual support for the Korean dataset. Furthermore, Bibek developed an end-to-end inference pipeline for paraphrasing. He used Pytorch, Matplotlib, transformers, and Huggingface models for his work. The replicated paper and the corresponding code can be found at the following links: Replicated Paper Link: https://tinyurl.com/qcpg-replication and GitHub Link: https://github.com/bibekyess/QCPG."
"Bibek participated in a project aimed at detecting intruders and notifying the house owner. He was responsible for setting up the hardware and managing the Node-RED workflow, connecting each process to the Telegram bot for an automated response. The tools used for this project were Node-RED. You can find a demo video of the project on YouTube: https://www.youtube.com/watch?v=YhccA0mm2YA and the code is available on GitHub: https://github.com/bibekyess/HomeSec."
"Bibek was part of a data cleanup project where he contributed to data visualization and webpage design. He used tools like Python, Plotly, Pandas, Flask, Dash, and Heroku for the project. A demo video of the project can be found at the link: https://tinyurl.com/data-anamoly-filter. The code for the project is available on GitHub at https://github.com/bibekyess/data-viz-project. The link for the project's live website is https://data-anamoly-filter.onrender.com. The objective of the project was to clean up the data and avoid the use of anomalies in downstream tasks."
"Bibek played a role in creating Project Planning & Management Document (PPMD), Software Architecture Document (SAD), and System Testing Document, where around 80% of the part in these documents was done by him. He wasn't explicitly involved in the development part, which was done by three other team members. Bibek helped resolve two issues in the project, one was converting the text to PDF and the other was managing dependencies using Conda and updating the README. Tools used included Django, React.js, NLP (Rule-based), Git, Notion, and Figma. The presentation link is https://tinyurl.com/lecsum-eosp and the GitHub link is https://github.com/NishantNepal1/LecSum."
"Regarding his experience of working with 6D Pose Estimation and Scene Graph Generation, he used Cosypose framework for 6D pose and Neural-Motif framework for Scene Graph Generation and performed experiments on custom cup datasets. This was done for a single-camera setup and he estimated 6D pose and generated dynamic scene graph. For 6 cups, with 6D pose estimation, relationship and attributes predictions, the combined framework ran at a speed of 1.25frames/s. Demo: https://tinyurl.com/pose-sgg-demo."
"Bibek did research on Probablistic Multiview Object Pose-Estimation. They proposed a novel probabilistic multiview object-pose estimation framework that 1) associates multiple estimation results with scene graphs, 2) combines pose distributions from single-view based estimators, and 3) construct a unified scene graph by predicting a unified pose distribution per object. They evaluate their method with table-top objectpose estimation scenarios and show the proposed approach outperforms three baseline methods in terms of position and orientation accuracies. His task involved conducting neural architecture research for pose estimation and scene graph generation, followed by training and performing combined efficient inference. Meanwhile, his friend was responsible for environment setup, creating a simulation dataset, and visualization. Draft paper link: https://tinyurl.com/kroc-draft."
"I have worked on AI-empowered F1/10 Autonomous Racing Car. Since I am more comfortable using Linux, I was in charge of setting up the environment and internet on the Jetson Xavier NX, including installing libraries and ROS packages for sensor interfaces. The other team members handled the hardware setup for the car, including the Arduino, LIDAR, IMU, camera, chassis, and wiring. My main responsibility was developing the perceptron, specifically object detection. I used Yolact_ros (it is instance segmentation-based framework and is more accurate) as we had plan to extend 2D detections to 3D detections for obstacle avoidance. However, it was computationally heavy, so we tried obstacle avoidance from LIDAR data. Later, due to computational issues, I used a simple rule-based classifier that involved cropping, masking, and counting the pixel values of red, green, and yellow. We only used the vision model for traffic light detection. For localization, we used AMCL and Hector Odom, while for control, we used PID. I also helped the team with algorithm testing and parameter tuning in the real-world environment. Tools Used: Yolact_ros, OpenCV, Linux, ROS, NVIDIA SDK, Arduino IDE Final Report: https://tinyurl.com/ee405-finalreport"
"I have worked on Movie Recommendation (Similar to NetFlix Challenge, 2006). My task was to fill in the <RATING> column of an input file that contained <USER ID>, <MOVIE ID>, and <TIMESTAMP> entries. To estimate the blank entries in the Utility matrix, which represented n users and m movies with rating values as entries, I used UVdecomposition for dimensionality reduction. This method approximates the matrix as R β‰ˆ UV^T, where U is an n x k matrix and V is an m x k matrix, with k being smaller than n and m. Each user and movie is represented by a k-dimensional vector of latent factors. To train the UV decomposition, I used an iterative approach called Alternating Least Squares (ALS) to minimize the root-mean-square error (RMSE). Before training, I performed two-fold normalization by subtracting the average rating of the user and then the average rating of the movie. Finally, I denormalized the utility matrix and used the U and V matrices to predict the rating for each corresponding user and movie entry. Tools Used: Numpy, Python"
"I have worked on the famous Pintos Project. It was the course project for my Operating Systems course, where I implemented pintos-kaist, a simple operating system framework for the x86-64 architecture. This project was forked from the pintos project at Stanford University, and we ran the code under the QEMU simulation environment. In the first project, my task was to solve thread synchronization problems with interrupts using semaphores, locks, and condition variables. In project 2, I implemented a system call handler for I/O interactivity in running user programs. In project 3, I implemented virtual memory management, including page fault handling and swapping, to build an illusion of infinite memory. In the final project, I worked on implementing the file system and succeeded on making the basic file system part to support addition, deletion, and manipulation of files. Overall, it was a great experience working in a large repository. Project Reference: https://casys-kaist.github.io/pintos-kaist/"
"The best way to get in touch with Mr. Bibek is by sending him an email at info@kc-bibek.com.np or by giving him a call/text at +82 (010) 4881-2332. Bibek is usually available during business hours and will do his best to respond to your inquiries as soon as possible. Here's a link to Bibek's resume for your convenience: https://kc-bibek.com.np/Bibek_CV.pdf"
"Regarding girlfriend and relationships, please contact Bibek directly. As his assistant, I cannot say these informations."