Srija Uprety

Hello, I'm Srija Uprety, a passionate enthusiast in the field of machine learning. I invite you to explore my website, dedicated to showcasing my journey and discoveries in the exciting world of artificial intelligence. Within this website, you will find a collection of my notable projects in machine learning and deep learning, highlighting the creative applications and the insights gained from each endeavor.

RESUME

EXPERIENCE

EDUCATION

fusemachines ai fellowship

Kathmandu, Nepal
2022-2023

Graduated from the fellowship program with an honors degree

Trinity INTERNATIONAL COLLEGE

Dillibazar Height, Kathmandu
2019-2023

       Bachelors degree in Computer Science and Information Technology with 80%(3.7 GPA) aggregate.

F1Soft International

2023-present
F1SOFT Group Tower, Lalitpur


Associate AI Engineer

Developing and implementing computer vision algorithms and models for various applications, such as object classification, optical character recognition, and image segmentation.
• Evaluating and fine-tuning existing computer vision models for imporving performance and accuracy.
• Designing and implementing efficient algorithms for real-time and embedded computer vision applications.

view full resume   Home

SKILLS

Python

Docker

Computer Vision

Machine Learning

Deep Learning

  Home

PROJECTS AN​D RESEARCH

MODELING MORPHOGENESIS IN 3D: NEURAL CELLULAR AUTOMATA

The use of Neural Cellular Automata (NCAs) has been proven useful in simulating the morphogenetic process. NCAs has found its application primarily in the 2D domain, however, this project expands the use of NCAs to the 3D domain by incorporating 3D convolutions into the neural network architecture. This would enable the simulation of complex 3D structures and spatial dependencies in all three dimensions. The project create a CA model that defines global coordination out of local level interactions. The research findings highlight two significant contributions, namely, an expansion of NCA to 3D voxels and the development of a cellular automation technique for producing voxel structures with different levels of complexity. The project explores multiple aspects, including the structural decay of the system after running for additional iterations beyond the training phase. It also showcases the regeneration property of the system. Additionally, the training behavior of the model was analyzed by varying the number of channels in different datasets. 

Keywords: Convolutional Neural Network Neural Cellular Automata, voxels, morphogenesis, Stochastic update, activation function, gradient optimization

  Home

Evaluation of Hybrid models to Estimate Forecasting Accuracy of Daily Global Solar Radiation: A case study of Parbat, Nepal


The demand for energy is increasing along with the population of the  world and the determination of a suitable source is a challenge as well  as an opportunity for modern technology. Solar energy is Nepal's most  important and sustainable energy resource. The high cost of GSR  measurement tools installation and calibration with high skilled  technical manpower has halted developing countries like Nepal. In this  study, different models are used for the prediction of solar radiation  potential. The dataset was obtained from an online database for the  location of Kushma, Parbat for ​the duration between 1992 to 2014.  Three standalone models ARIMA, ANN, and LSTM, and two hybrid  models ARIMA-ANN and ARIMA-LSTM were...   .   

Read More    Home

OTHER PROJECTS

Time Series Analysis

The major objective of this project was to enhance the deep learning models that could potentially perform well in prediction problems like value of stocks.

learn more

Image to text converter

The project aims to convert given image into text that can be copied and modified.

learn more

Correlation between reviews and star ratings

The main purpose of this project is to discover if there is a correlation between reviews and star ratings.

learn more
  Home

Thank you for your interest in my research. Please note that my research is currently undergoing rigorous evaluation and review.

Updates regarding the evaluation progress and anticipated publication date will be shared as soon as they become available. Thank you for your patience and understanding.

Go Back