Professional Documents
Culture Documents
EMPLOYMENT
Fusemachines (Sept 2019 - Present)
ML Engineer Level 2
(Jan 2022 - Present)
ML Engineer Level 1
(March 2021 - Dec 2021)
● Recommendation engine and web development for a software platform of a food company.
Devised a recommendation engine for diet based ingredient substitution based on ingredient data and user
preferences. Also developed and updated various APIs for product management in the web application.
Techniques used: Jaccard distance, cosine distance, word2vec
Tools used: Flask-restplus, MongoDB, Kubernetes, Google Cloud Platform, Mongo Atlas, numpy, pandas
Associate ML Engineer
(Sept 2019 - Feb 2021)
● Developed a system to detect and extract tabular data from PDFs (text/image based).
Trained a Faster RCNN model to detect tables in pdf images and tried various third party applications to
extract tables from the detected regions.
Also used classical image processing techniques (eroding, dilation, histograms, Gaussian kernels) to create
table structure.
Techniques used: Object detection, text processing (lemmatization, POS, NER)
Tools used: CNN, spacy, pdfminer, scipy
● Developed a prototype system to extract tabular data from lower quality images (image based).
Used super-resolution techniques combined with Tesseract-OCR to extract data from invoice images.
Techniques used: Super-resolution, OCR, Classical image processing
Tools used: CNN, scipy, Tesseract-OCR.
EDUCATION
Pulchowk, Nepal Tribhuwan University 2015 - 2019
● Bachelors of Engineering in Computer Engineering
Coursework: Software Engineering; Computer Architecture; Data Structure and Algorithms; Artificial
Intelligence; Database Foundation; Operating Systems; Calculus; C; C++;
TECHNICAL EXPERIENCE
Projects
● Latent variable analysis of StyleGAN: Analysis of relationship between the input random vectors of StyleGAN
and the features of corresponding generated facial images which enables generation of facial images with
desired features.
● Electronic Health Record (EHR) system to store medical records using Django.