Project Portfolio
Project 1: Voice AI Agents for Customer Service | Industry: Retail/E-commerce
● Scope: Implement voice-based AI assistants to handle incoming customer
service conversations.
● Responsibilities:
o Designed and deployed voice AI agents for autonomous customer
interactions using speech to text and text to speech models.
o Integrated agents with internal systems and external tools for seamless
operation.
o Optimized conversation workflows to improve response accuracy and
efficiency.
● Environment: Python, CrewAI, VAPI, n8n, Streamlit,
Twilio, Azure OpenAI, Azure STT, Azure TTS
Project 2: Scalable and Robust ML Pipelines for Multiple Use
Cases | Industry: Manufacturing & E-mobility
● Scope: Implement end-to-end ML pipelines for multiple use cases,
ensuring scalability, robustness, and CI/CD integration.
● Responsibilities:
○ Designed and implemented dynamic ML pipelines for multiple use cases
with dynamic parameterization.
○ Built model registry for versioning and deployment tracking.
○ Consolidated datasets for each use case, ensuring clean inputs.
○ Integrated pipelines with CI/CD infrastructure for smooth deployment.
● Environment: Python, Snowflake, Snowflake Model
Registry, GitHub Actions, CI/CD, MLOps
Project 3: Automated KPI Extraction from Documents |
Industry: Retail/E-commerce & Manufacturing
● Scope: Develop automated pipelines to extract KPIs for various client
and prospect documents from diverse file formats efficiently.
● Responsibilities:
○ Implemented KPI extraction pipelines with dynamic control outside of
code with flexible configurations.
○ Developed Streamlit dashboards to visualize pipeline outputs and
dashboards for monitoring and reporting.
○ Mentored team members on pipeline development and dashboard
integration.
● Environment: Python, Streamlit, Snowflake, Snowflake Cortex, Azure
OpenAI, CI/CD, Data Pipelines
LPDG India Pvt. Ltd. | Managing Consultant (Big Data &
Data Science) - Senior Data Scientist Apr’23 - Apr’25
Project 1: Development of Multiple Custom Data Conversational Chatbots
Using GPT-3.5 and Azure OpenAI
● Scope: Develop 7 RAG-LLM-based custom data conversational chatbots
using GPT-3.5 with Streamlit as the interface using Azure OpenAI.
● Responsibilities:
○ Integrate GPT-3.5 for sophisticated natural language processing to
enhance user interaction.
○ Design and build chatbots equipped with advanced feedback
mechanisms.
○ Integrate suggested follow-up questions to enhance interactive
capabilities.
○ Implement support for multiple languages to ensure broad accessibility.
○ Leveraged Langchain’s Conversation Retrieval Agents and
Conversation Memories to integrate custom prompts based on client
requirements with contextual conversation history.
○ Leveraged FAISS from Meta and Cognitive Search from Azure for
Vectorstore.
○ Integrated Docker and Kubernetes for efficient deployment of
machine learning models, automating infrastructure management and
scaling using Terraform.
○ Utilized AWS (S3, Lambda) to manage and store chatbot data and CI/CD
with GitHub Actions.
○ Integrated FastAPI as the API framework to handle chat requests with
Azure OpenAI, enabling efficient and scalable API endpoints for real-time
conversation data and feedback handling.
○ Communicated directly with clients concerning deliverables,
requirements, and feedback.
○ Supervised and mentored a team of 4, fostering a collaborative and
productive work environment.
● Environment: Python, Snowflake, Langchain, Azure, OpenAI, docker,
Streamlit, Prompt Engineering, SQL, Generative AI, FastAPI
Project 2: Development of Applications for Invoice Data Extraction and Rule-
Based Categorization
● Scope: Develop 3 RAG-LLM-based Streamlit applications to fetch
information from invoices and categorize rules.
● Responsibilities:
○ Extract relevant data from various invoices using LLMs by
customizing Langchain’s Retrieval Agents and their prompts to identify
different fields from the documents.
○ Store and manage extracted data in the Snowflake database for efficient
processing.
○ Implement rule-based categorization to classify and organize invoice
data.
○ Implemented Jenkins for automated CI/CD pipelines, integrating
Docker containers for application deployment.
○ Used Terraform for infrastructure-as-code, automating AWS resources
for efficient application scaling.
○ Integrated FastAPI as a lightweight API layer to expose endpoints for
fetching, categorizing, and retrieving invoice data.
○ Employed FastAPI’s async capabilities to handle high concurrency,
improving the application's response time for large datasets.
○ Ensure data accuracy and consistency through rigorous validation and
testing.
○ Optimize application performance to handle large volumes of data
efficiently.
○ Engaged directly with clients to discuss deliverables, gather
requirements, and incorporate feedback.
○ Managed a team of 3, delegating tasks and responsibilities to optimize
team performance and project outcomes.
● Environment: Python, Snowflake, Langchain, Azure, OpenAI, docker,
Streamlit, Prompt Engineering, SQL, Generative AI, FastAPI
Project 3: Creation of a Dashboard for Converting Unstructured Text into
Structured Output for Emails
● Scope: Develop a RAG-LLM-based dashboard to convert unstructured
text into structured output for formatted emails.
● Responsibilities:
○ Integrate the dashboard with GPT-3.5 and Azure OpenAI for advanced
text processing.
○ Leveraged Langchain’s Agents to provide the output email template.
○ Design a user-friendly interface for easy input and output of data.
○ Automate the formatting of emails based on structured data to improve
communication efficiency.
○ Utilized FastAPI’s support for request validation.
○ Leveraged Apache Airflow for task orchestration and scheduling to
automate the dashboard's data pipelines.
○ Used Apache Spark on AWS EMR for efficient large-scale data
processing.
○ Provided user support and troubleshoot issues promptly.
○ Collaborated directly with clients and business stakeholders to
discuss deliverables, gather requirements, and incorporate feedback.
● Environment: Python, Snowflake, Langchain, Azure, OpenAI, docker,
Streamlit, Prompt Engineering, SQL, Generative AI, FastAPI
Project 4: Design and Implementation of a Plug-and-Play Environment for
Custom Chatbots on Azure OpenAI
● Scope: Design a plug-and-play environment for custom chatbots on
Azure OpenAI integrated with Snowflake.
● Responsibilities:
○ Create scalable and customizable chatbot solutions that can easily
adapt to different client needs.
○ Utilized Kubernetes to manage and orchestrate chatbot deployment,
ensuring scalability and reliability.
○ Created RESTful APIs using FastAPI to allow seamless integration of
custom chatbot components.
○ Automated deployment processes with Jenkins and GitHub Actions,
using Terraform for infrastructure provisioning.
○ Integrate chatbots with Snowflake for efficient data management and
processing.
○ Provide comprehensive documentation and training to ensure users
can effectively utilize the chatbots.
● Environment: Python, Snowflake, Langchain, Azure, OpenAI, docker,
Streamlit, Generative AI, FastAPI
Project 5: Development of a Real-Time Human Shape and Gesture Tracking
Application Using Computer Vision
● Scope: Develop a computer vision application for real-time human
shape and gesture tracking.
● Responsibilities:
○ Design and implement advanced gesture recognition algorithms to
track human movements accurately.
○ Develop a real-time tracking interface to display and analyze gestures.
○ Present the application at WETEX Dubai, showcasing its capabilities and
potential applications.
● Environment: Python, Streamlit, openCV, SQL
Project 6: Implementation of an Outlier and Anomaly Detection Pipeline with
Multiple Machine Learning Algorithms
● Scope: Implement an outlier and anomaly detection pipeline with
multiple ML algorithms.
● Responsibilities:
○ Develop and test over 10+ machine learning algorithms to identify
outliers and anomalies.
○ Handle class imbalance through 15+ different approaches to ensure
accurate detection.
○ Integrated AWS services (S3, Lambda, SageMaker) to automate the
machine learning pipeline.
○ Used Docker for containerization and Jenkins for automating the
deployment process.
○ Integrate the detection pipeline with existing systems for seamless
operation.
○ Document the pipeline and provide training to users for effective
utilization.
● Environment: Python, Jupyter notebooks, streamlit, scikit-learn, SQL
Vivere GmbH Hamburg,
Germany | Applied Data
Scientist May’22 - Feb’23
Project 1: Creation of an LLM-Based Intelligent Dashboard for Identifying New
Product Ideas
● Scope: Create an LLM-based intelligent dashboard to identify new
product ideas.
● Responsibilities:
○ Utilize Azure OpenAI to enhance product strategy and identify market
opportunities.
○ Develop the dashboard interface using Streamlit for intuitive user
interaction.
○ Integrate multiple data sources and analytical tools to provide
comprehensive insights.
○ Provide user training and support to help stakeholders leverage the
dashboard effectively.
● Environment: Python, Streamlit, Scikit-learn, PostgreSQL,
Google Bigquery, Azure OpenAI, Prompt Engineering, Generative AI
Project 2: Development of an Analytics Dashboard for Marketing Spending
Trends and Sales Analysis
● Scope: Develop an analytics dashboard for marketing spending trends
and sales analysis.
● Responsibilities:
○ Analyze marketing spending and sales data to identify trends and
insights.
○ Develop visualizations to show spending trends and sales performance
clearly.
○ Contrast organic vs. inorganic sales to provide a detailed analysis.
○ Optimize dashboard performance to handle large datasets efficiently.
○ Provide actionable insights and recommendations to stakeholders
based on the analysis.
○ Leveraged FastAPI’s Pydantic models for data validation, ensuring
the accuracy and integrity of the incoming analytics data.
● Environment: Python, Streamlit, Pandas, scikit-learn, SQL, FastAPI
Project 3: Implementation of a Competitor Analysis Dashboard Using
Statistical Analysis and Clustering
● Scope: Implement a dashboard to identify competitors using statistical
analysis and clustering.
● Responsibilities:
○ Analyze competitor data across various marketplaces to identify key
competitors.
○ Develop clustering algorithms to categorize and group competitors.
○ Create visualizations to display the competitor landscape and insights.
● Environment: Python, Streamlit, Pandas, scikit-learn, SQL
Project 4: Engineering an Event-Based Notification Service for Email and
Asana
● Scope: Engineer an event-based email/Asana notification service.
● Responsibilities:
○ Develop notification triggers to alert users of significant product
changes and updates.
○ Integrate the notification system with email and Asana for seamless
communication.
○ Utilized FastAPI to handle API calls for triggering notifications,
integrating with external services like email servers and Asana APIs for real-
time alerts.
● Environment: Python, Email/Asana Integration, FastAPI
IBMI Universitätsklinikum, Otto-von-Guericke Universität (OvGU)
Magdeburg, Germany | Machine Learning Engineer
Oct’20 - Mar’22
Project 1: Development of a Human Brain Model (VisMem) to Perform Serial
Recall Tasks
● Scope: Develop a human brain model (VisMem) to perform serial recall
tasks that identify visual inputs from digits similar to human eyes as input
sources.
● Responsibilities:
○ Conduct extensive research on cognitive brain models and their
applications.
○ Implement and test the functionalities of the brain model to ensure
accuracy.
○ Analyze the performance of the brain model in various cognitive tasks.
○ Document research findings and methodologies for future reference.
○ Collaborate with the research team to improve and refine the brain
model based on experimental results.
● Environment: Python, Nengo, Tensorflow, openCV
Project 2: Research on the Effects of Drugs and Injuries on Brain
Performance
● Scope: Research the effects of drugs and injuries on brain
performance while the visual input sources are not up to 100% efficiency.
● Responsibilities:
○ Analyze the performance of cognitive tasks under different
conditions to understand the impact of drugs and injuries.
○ Develop methodologies for testing and evaluating brain performance.
● Environment: Python, Nengo, Tensorflow, openCV
Project 3: Comparison of Query-By-Committee (QBC) and Single Learner
Uncertainty for Medical Image Segmentation on Brats2017 Dataset
● Scope: Investigated the performance of uncertainty metrics in medical
image segmentation, comparing Query-By-Committee with Single Learner
Uncertainty.
● Responsibilities:
○ Explored uncertainty metrics like KLD, JSD, and Entropy to improve
informative sample selection by 7% over random sampling for an active
learner.
○ Demonstrated that Query-By-Committee provides better
informative samples with only a 1% fluctuation in precision compared
to a 12% fluctuation with single learners.
● Environment: Python3, PyTorch, Nibabel, NumPy, SciPy, Matplotlib,
MedicalZooPyTorch, modAL, skorch, UNET, VNET
Project 4: Interpretable Deep Reinforcement Learning (eXplainable AI) on
MNIST and Space Invaders
● Scope: Developed an interpretable deep reinforcement learning
model to improve explainability in complex vision-based tasks, focusing on
attention visualization in image-based decision-making processes.
● Responsibilities:
○ Applied deep reinforcement learning (DQN) to computer vision tasks
such as MNIST digit classification and Space Invaders, enabling the model to
make decisions based on visual input.
○ Enhanced model interpretability by generating saliency maps that
highlight regions of interest within images, showing where the model
focuses during decision-making.
Experience 2 Years - 3 Years
Work Style Remote
Duration Within a Week
Location Abhayapuri