About
In 2019, I worked on a deep learning project to accelerate Topology Optimization for Automobile Body Parts and have been working on AI projects ever since.
My focus these days is deep-diving into LLMs and building Autonomous Multi-Agent Systems for SaaS products at Dbiz.AI.
In my free time I do data science projects to study the stock market. I also spend time playing around with diffusion models.
Experience
-
April '22 — Sep '23 ➢ Lead the chatbot engineering team to build & maintain NLP powered chatbots for our portfolio SaaS software - onboarded 5 new clients
➢ Developed the chatbot and API layer for the new Candidate Relationship Management (CRM) SaaS product - faster feature deployments using new architecture, reducing client onboarding times by a factor of 3
➢ Initiated performance monitoring procedures for all client chatbots to promote responsible AI practices and improve answering accuracy from 61% up to 89%
-
Python
-
Flask REST API
-
Shell
-
MongoDB
-
Natural Languge Processing
-
Microsoft Azure
-
Kubernetes
-
Jenkins CI
-
Docker
-
SQL
-
Git
-
RASA
-
Jira
-
-
Jan '21 — April '22 ➢ Worked in Product Development - coordinated digital design, validation & testing, prepared DFMEA reports
➢ Analysed issues with supplier contributions based on customer feedback to detect defects & report findings
-
Product development
-
Supplier management
-
Customer centricity
-
VBScript
-
MS Excel
-
-
July '19 — June '20 ➢ Developed & trained a novel neural network model to predict the solution of a conventional 2D Topology Optimization problem
➢ Posed as an image segmentation problem, a U-Net based solution methodology was proposed
➢ ResNet modules used in encoder-decoder networks to achieve accuracy (93.7 % in test dataset) higher than previously published models
-
Python
-
Computer Vision
-
SQL
-
Scikit-learn
-
Pandas
-
NumPy
-
Pytorch
-
VBScript
-
Projects
-
Intelligent Virtual Assistant with NLU for CRM - SaaS product
➢ Developed a chatbot from scratch using RASA framework for candidates to raise concerns & answer queries using NLU, REST API layer built using Flask to handle business requirements
➢ MongoDB used to store chat history data, application deployed to Kubernetes cluster using Jenkins CI
➢ The new virtual assistant architecture enabled faster feature shipments - reducing client onboarding times by a factor of 3
-
Atliq Technologies' Sales Data Analysis & Visualization using MySql, PowerBI
➢ The real sales data of Atliq Technologies was cloned from CodeBasics github repository (made publicly available for academic purposes)
➢ MySql was used for data wrangling and analysis, PowerBI Desktop was used to create a visualisation dashboard
➢ Insights were drawn on zone-wise sales all over India and identify top products, sales/profit margins in different cities
-
Introduction to Diffusion Models
➢ A collection of jupyter notebook tutorials created based on "How Diffusion Models Work" - a short course by DeepLearning.AI
➢ Added a new data visualization notebook for better understanding the data, fixed a few bugs and added new functions to facilitate easier follow-up of the original course