Jun 2022 – Aug 2024 · 2 yrs 2 mos

• Engineered, modeled and assembled processes and concepts within finance. This encompasses re-engineering PnL reporting system to streamline the daily operations.
• Developed a personalized state-of-the-art data model which reconciles data from various streams to store them as a single master table which is utilized in various day to day operations.
• Managed manual deployments on Inhouse-Deploy platform to ensure punctual and successful releases with no failures.

Jan 2022 – Jun 2022 · 5 mos

• Worked on Daily-Submission reject WebApp where I engineered scattered streams of data into a standardized Timeseries framework.
• Utilized Enaml to create an interactive webapp which aims to decommission the PowerBi end user application which in-turn eliminated significant cost
• Planned, demonstrated and successfully led production launch of the project which was then leveraged by several teams to be re-engineered to their own book of work and set of problems.

Jun 2019 – Jul 2019 · 1 mo

• Aim: To create an AI-powered system for detecting Motorcycle drivers not wearing a helmet.
• Summary: We developed an AI system following the then-recent traffic law making Helmet a compulsory accessory for Motorcycle riders. The goal of the system was to detect the perpetrators, save their details and make the information accessible for further investigation.
• Technical Details:
• ⚫ Data Source: There is no readily available Image dataset for such a task. The sourcing of data was done by capturing images of various traffic footages every 5 seconds.
• ⚫Preprocessing of data: Labelimg software was used to individually annotate and draw bounding boxes to form a training dataset
• ⚫AI model: State of the art model was fine-tuned to perform the desired operation.
• ⚫ Unique Features:
• 1)The system is capable of saving images of the detected subjects in a folder.
• 2)System can also distinguish between helmets and scarfs which usually confuse many such models.