Firdous Sultana

I believe my strongest contributions are in generating creative solutions for various problems.
I like to break complex tasks into smaller subproblems for a thorough analysis
and then solve them impactfully.
I am a quick learner and I like to take challenges. Like many developers, I like to design and execute
small projects on the side that use a tool or technique in order for me to learn it through applying.
I have a habit of learning something new every day and then trying to apply that in practical fields.

I am also an awesome team player. I am really very gracefull for having the ability to manage
a team. My team " blackk_coderss " won Smart India Hackathon 2019 for solving a real-time
problem statement.
I love learning new things which is why I love being on a team.
For every group project we accomplished,
we had a half dozen different ideas and ways to tackle it.
We all worked together to refine those into one final plan.
Being open to what others have to say and seeing new and exciting ways others would think
outside the box not only keeps me interested in the work but drives me to excel
outside of work so I could continue to contribute to the group overall.

I find Machine Learning endlessly fascinating.
The most complex project that I have worked upon so far was during my internship at Acrolinx
GmbH, Berlin, Germany. I was responsible for "Development of automatic recognition of missing articles in English sentences" using an existing Neural Network model in Python an Jupyter Notebook. It was a really challenging experience for me to solve this problem of 'Deep Learning'.
I have been studying Machine Learning from my sophomore year of undergrad and making various projects on it. However, having a sudden shift from ML to DL, for solving such a complex task in a small span of time was a huge opportunity to prove myself! The "Missing Articles Problem" was totally new for me. It was the first time I encountered the problem of 'Deep Learning' and yet I was still able to do it! The main challenge that I faced was that "Missing Articles Problems" falls between grammatical use and the different style of using articles in different areas.
The non-availability of a proper, annotated dataset made it even more intricate.
I applied various techniques of 'Natural Language Processing' under Deep Leaning, like using biLSTM model, word2vec model, etc. to solve the task. The most intriguing one was Google's pre-trained "BERT" model that helped in increasing the accuracy of the problem.

I was highly illuminated by the knowledge of NLP and DL after solving this problem.
I developed the habit of reading scientific research papers independently, which in turn helped in the uplifting of my understanding capabilities. While working on the project, I started to learn from my mistakes and corrected them.
As being completely new to practical, corporate world setting, every hour spent at Acrolinx
gave me some amount of experience all the time, all of which cannot be explained in words.
I gained valuable hands-on Deep Learning experience through my engagements in various real-world projects
and I aspire to work more.