Nov 2024 – Present · 1 yr 4 mos · Beijing, China · On-site

• Working at TSAIL lab under supervision of Prof Hang Su

Apr 2024 – Sep 2024 · 5 mos · On-site

Aug 2022 – Mar 2023 · 7 mos · San Francisco Bay Area · Remote

• # 2 hire for the company, handled almost every part of engineering ( mostly worked on Machine Learning, Backend, DevOps, and Infrastructure domain)
• Developed microservices in Golang to handle dataset import, model training, on-device
• optimization, and model deployment tasks for end users without requiring them to write any code
• ・ Implemented a pipeline to automatically deploy the microservices and frontend applications in
• Golang, NodeJS, Python, and ReactJS using Github Actions
• ・ Worked on an orchestrator service that can launch and terminate GPU instances depending upon
• the load on various Cloud Service Providers for model training
• ・ Developed Data import background job that can import Image Classification and Object Detection
• datasets in multiple formats from an archived file upload
• ・ Implemented a data versioning system where users can store a single dataset in multiple versions
• ・ Developed APIs to display dataset statistics to the end user using aggregations
• ・ Developed backend for data splitting feature through which users can create random splits (train,
• test, val) or move samples from one split to another using a simple Graphical User Interface
• ・ Developed a pipeline that allows users to train Image Classification and Object Detection models
• with real-time monitoring by just selecting the dataset and setting values for hyperparams through
• simple web GUI
• ・Developed an iterative pruning algorithm that can prune a trained model on an incremental basis
• until the performance of the model does not drop below the specified threshold.
• ・Developed a pipeline to export trained models to different formats(like ONNX, OpenVINO, CoreML,
• etc.) for deploying those trained models on edge devices

Oct 2021 – Jul 2022 · 9 mos · Kathmandu, Bāgmatī, Nepal · Remote

• Involved in building end-to-end ML pipelines for extracting structured information from unstructured documents using the blend of Computer Vision and NLP techniques:
• ・Worked on Template free extraction of key-value pairs in a document
• ・ Worked on error analysis and improving the performance of Handwriting Recognition Models for
• English and Nepali texts
• ・ Developed algorithm to detect and crop photographs of persons from the banking forms using
• classical Computer Vision techniques.
• Involved in designing and implementing software architectures to serve ML models in production:
• ・Database Design to store all the required information
• ・Dockerized microservices and deployed them using docker
• ・Designed and implemented workflows that allow modifying the results (if required) and verification
• of the results before using the results predicted by the model
• ・ Implemented an authentication and authorization system to assign different roles to different users
• ・ Scaling the application to handle multiple requests simultaneously
• Demoed the product to multiple potential customers which eventually led to a contract signing with a potential customer for further collaboration.

Jul 2021 – Jul 2021 · 0 mo · Los Angeles, California, United States · Remote

• ・ Led a pod of 10 students from 5 different countries for a 3-week Computational Neuroscience
• Summer School
• ・ Assisted the participants in the programming part of the project
• ・ Managing meetings with project mentors for project ideas discussion

Apr 2021 – Jun 2021 · 2 mos · Kathmandu, Bāgmatī, Nepal · Remote

• ●Developed fully responsive webpage from a psd template using HTML and CSS
• ● Developed various projects like Image Slider, Box Collision, Ant Smasher, Car Lane Game and Flappy Bird game using vanilla JS.
• ● Developed a "Mathematical Functions Visualizer”using plain Javascript and Canvas.
• ● Developed CRUD API using Node.JS

Dec 2019 – Mar 2020 · 3 mos · Kathmandu, Bāgmatī, Nepal · On-site

• Implemented various Computer Vision Algorithms in C++
• Training deep learning models using C++