Data scientist with expertise in NLP, computer vision and an
Hi,
I am Pritam, an expert data scientist of applied AI research with real experience in predictive analytics and commercial NLP and computer vision projects in python 3.6 and 3.7.
Other than truelancer.com freelance work, I have 1 year of ML and DL freelancing experience on 6 real AI projects for my upwork clients from USA, Canada, Japan and Spain with their happiness on my completed work.
Here is list of what I can do with expertise for you -
- DIVERSE TEXT PROCESSING,
document classification,
segmentation,
topic modeling,
intent extraction,
NER extraction,
NLU,
NLP models training from scratch and using transfer learning
and use state-of-the-art pre-trained vector embeddings such as
BERT, xlnet, elmo and glove models
for various type of NLP tasks.
- BUILD ANY TYPE OF CHATBOT USING GOOGLE DIALOGFLOW AND FIREBASE WITH GOOGLE ASSISTANT, GOOGLE HOME AND GOOGLE HOME MINI,
AMAZON ALEXA,
IBM WATSON,
MICROSOFT LUIS,
FACEBOOK MESSENGER,
RASANLU,
CHATTERBOT
AND NLTK.
- image and video
classification,
categorization,
segmentation,
tracking,
recognition,
identification,
watershed,
yolo v3 training from scratch or using as it is,
training image processing models from scratch and by using transfer learning
and using state-of-the-art pre-trained computer vision models such as
mxnet,
inception v4,
efficientnet,
alexnet,
resnet- 32, 50, 56, 101,
imagenet,
open images,
inception resnet v3
and vgg.
- on excel/csv of about 1 GB, do
data cleaning,
missing data changing
& wrangling
with
statistical & logical analysis of data and high accuracy prediction on data.
- hyperparameter tuning for deep learning models and
use automl like autokeras or
use AWS or Google cloud and azure
auto machine learning options for super hypertuning deep learning models.
- use
AWS, Google cloud and azure
for deep learning models training and optimization.
- use 16 types of evaluation metrics.
- build deep learning models with high predictive accuracy and performance.
- scrape web using scrapy for super fast, performance oriented and consistent internet sites data extraction and save the extracted data to csv/text/json/xml/pickle/hickle/hdf5 formats.
- save data science projects's data to csv, json, XML, text, pickle, hickle, hdf5 and save trained models to tflite for android applications and also save models to hdf5, ckpt and pb formats.
- data visualization using plotly, dash, d3.js by nvd3, ggplot by plotnine, matplotlib and seaborn, wordcloud python libraries.
- SQL and NoSql including SQL server, mysql, postgresql, mongodb and sqlite automation using python.
Also, I have very good experience on django, flask and nodejs development with python 3.6 and 3.7 including
- state-of-the-art and sophisticated API development
and
- backend and frontend connection with state-of-the-art dashboard
using these.
And I use python libraries name - scrapy and regex to scrape data from internet sites.
Currently I work in python 3.7 in anaconda jupyter ipynb notebook by
tensorflow,
keras,
pytorch,
scrapy,
regex,
spacy,
nltk,
textblob,
rasa,
opencv,
yolo v3,
sklearn,
django,
flask,
tflearn,
automl by autokeras,
pyspark,
MLib
and lot of other python libraries.
Links and descriptions of my kaggle and github AI projects on ML, DL, RL, big data and internet data scraping are in portfolio. You can check these.
Also, I worked on huge datasets such as csv file of 4 million rows.
I also have experience on reinforcement learning and big data.
Looking forward for your data science project....
