May 2021 – Mar 2024 · 2 yrs 10 mos · Silicon Valley, California, United States

• Doorstead, A US based startup (Series A 16M$ ) offering a property management plattform for homeowners. I worked in an agile team of 6. As a Data Scientist I built statistical and machine learning models (CoxPH) using Pandas, Numpy and AWS to predict the number of days taken before a property is rented out considering a variety of factors. The model is one of the USP’s of the company since it helps to understand after how many days the company will make money on a particular property.
• I also developed an algorithm that shows similar properties to a user-defined target property using Pandas, Numpy & AWS. Used FAISS algorithm, to search for similar data points in hyperdimension metric space (similar to KNN). Also, created a similarity_score metric to rate resultant properties on the basis of their similarity and plot them on the dashboard. It reduced the effort and search time of a pricing analyst by more than 80%. Technologies used: Python | AWS | Docker | Linux | Statistical Modeling | Machine Learning | Algorithms | Streamlit | Time Series| MySQL | Pandas | Numpy | Jira

Mar 2021 – Apr 2023 · 2 yrs 1 mo

• I love sharing knowledge and joined Datacouch as an ML Intructor in to educate in my free time

Apr 2020 – May 2021 · 1 yr 1 mo · Mumbai, Maharashtra, India

• Project: Arya.ai is an Indian-based startup offering an autonomous AI operating platform used by 20+ banks worldwide. I worked in a team of 5 developing different core features of Arya. I built and trained a deep learning model using Tensorflow & Keras which scans and understands the handwritten and printed text. It extracts the information in the image using NLP, OCR, Computer VIsion and saves it in MySQL. Deployed the model in the bank's private cloud network. The model was first used by a big client where it processed 500K+ cheques daily and was later deployed to other clients as well.
• I also developed a model to process image-based invoices to extract tables and table-like structures using Computer Vision, OCR, and NLP to create a CSV from image-based invoices. I trained Tablet( Table Neural Network) on custom-labeled invoices to detect table structure and columns. Used an unsupervised approach to detect rows and finally used OCR + NLP techniques to properly extract the text and save it inappropriate rows and columns.

Oct 2017 – Apr 2020 · 2 yrs 6 mos · Noida Area, India

• Built a Batch Processing pipeline of sensor data using Apache Spark and Python.
• Build API for various stakeholders of the project and deployed on Linux servers.
• Built an autonomous Anomaly Detection framework using a mix of computer vision, filtering, and statistical techniques.
• Managed the entire Data Warehouse built on top of HDFS.
• Refactored the entire code base of the end to end machine learning pipelines.
• Used Fully connected Convolutional Neural Network for Image segmentation using U-Net.
• Lots of Learnings from the project :
• CRISP-DM Process: How each and every component plays a vital role in the final results.
• Effective Communication is the key when you are solving problems for businesses. Nobody will be able to tell you everything at the first meeting. It is our responsibility as a Data Scientist to ask the right kind of questions to the businesses. Data Science is more than just problem solving, it is first understanding the problem deeply and fundamentally.
• How a research team works. We made custom tools to measure the metrics that we were chasing. We made algorithms to accomplish our goals. Everything was made from scratch. It is soo much fun for Maths Enthusiasts to live in an environment where you can solve problems through Maths. I realized how important it is to feed our soul with the satisfaction of good work.
• Learnt Lifecycle management of models. How to calibrate models and update them as and when required.

Jun 2016 – Oct 2017 · 1 yr 4 mos · New Delhi Area, India

• Yourslate is an e-commerce platform to buy portraits for your pets, friends, and family. The basic idea was to make it easy for anyone who wants to transform their memories into portraits and put them on the wall.
• Operation :
• We operated for a period of over 18 months. We launched a basic website wherein you could place an order selecting your preferred frame, size, and artist also. We provided art in charcoal, water, and oil on canvas.
• We won a national level competition in Bengaluru for prize money of 5 lakh and later got funded through the event.
• We started international works for the US, Europe, UAE, and Australia after 8 months of our operation and scaled aggressively from there.
• Over a period of 18 months, we scaled to 4 continents, over 5 core team members, and a revenue of over 50,000$.
• Learnings
• Founders are Salesperson at heart.
• Leverage is how this world works. Some people also call it to help.
• Identify the right channels for your sales funnel and keep exploring more channels.
• Make fewer partnerships and more sales.
• Keep your team happpy and motivated.