Kranti Kumar Punyamanthula

+91-9502955443 | kranthi55@rocketmail.com

Profile summary
Results-driven Data Scientist with 5+ years of experience and strong background in machine learning, deep learning, and
analytics. Skilled at transforming raw data into actionable insights that influence business strategy and drive innovation.
Committed to continuous learning and enhancing organizational success through advanced data solutions.

Skills
Machine Learning: Decision tree, XGBoost, RandomForests, Support Vector Machine (SVM), K-Means Clustering,
Model Building, MLops
Deep Learning: Neural Networks (ANN), Convolution Neural Network (CNN), LSTM, NLP/CoreNLP, Computer Vision
Object Detection and Segmentation, Recommendation systems, AutoEncoders, Generative adversarial Networks
Generative AI: Large Language Model (LLM), BERT, Fine-Tuning LLMS, Transformers, GPT, Encoder-Decoder
Models, RAG
Cloud Tech: AWS EC2, AWS Bedrock, AWS Lambda, AWS SageMaker, AWS S3 Bucket, AWS Cloudwatch, AWS
Rekognition
Libraries: Scikit-learn, TensorFlow, PyTorch, Keras, JAX, NumPy, Pandas, Matplotlib, Keras, Spacy, OpenCV, Seaborn
Programming Languages: Python, MySQL, PHP, C
Databases: Hive, Vertica
Others: Tableau, Hugging Face, Anaconda, Prompt Engineering, Advanced Excel, Git Version Control, Selenium
LLMS Used: ChatGPT, Llama 3, Mistral, Anthropic Claude, Titan Lite, LLava multimodal

Experience
Matrimony.com May 2022 – Present
Data Scientist manager Chennai,IN
• Led and deployed multiple machine learning projects across the organization, ensuring successful
implementation and ownership.
• Conducted statistical modeling to uncover hidden patterns and trends in customer activity data across various

products, including Jodi FUP limits, retail analysis and various other products.
• Wrote complex SQL queries using Hive, Vertica, and Hadoop, conducting analyses to assess new product

changes and their impacts, providing valuable insights to support business development strategies for internal
stakeholders, including marketing and product engineering teams.
• Engaged with internal stakeholders and department vice presidents to identify opportunities for growth and

improvement, fostering a data-driven culture throughout the organization.

• Built Tableau reports to visualize and analyze data trends, enabling stakeholders to make informed, data-driven
decisions.
• Engaged regularly with department vice presidents and cross-functional teams to understand project

requirements, share updates, and facilitate timely deployments, fostering alignment and collaboration across the
organization.
• Ensured data integrity and accuracy by working closely with data engineering teams to perform data quality

assessments and data cleansing processes. Led initiatives to maintain and enhance database accuracy, ensuring
high-quality, accessible data for analytics and reporting purposes.
• Leveraged MLOps practices to build and deploy robust machine learning pipelines, automating workflows and

ensuring scalability, reproducibility, and effective model monitoring for continuous improvements.

VinAudit.com (Acquired Simple Intelligence) June 2021– May 2022

Data Scientist Remote
• Extracted and developed parser scripts to efficiently load millions of records using PHP, Selenium, and Python.

• Developed and implemented a Generative Adversarial Network (GAN) model for artificial vehicle image synthesis,
 utilizing millions of images for training.
• Developed and implemented a Generative Adversarial Network (GAN) model for artificial vehicle image synthesis, utilizing
millions of images for training.
Simple Intelligence Jan 2020– June 2021
Machine learning engineer Remote
• Conducted research on relevant papers and developed proofs of concept (POCs) for image segmentation and

Generative modelling.
• Worked as an Individual contributor

• Coordinated and worked with CEO in an early startup agile environment working from choosing hardware and setting up

machines from scratch.

Projects
Photo Validation | Python, CNN, AWS Rekognition, REST Api, AWS Lambda, Computer Vision, MLOPS
• Developed and Deployed Photo Validation code to automate photo validation process leveraging AWS

Rekognition and AWS Lambda services.

• Implemented an image classification model using RESNET 50 model to reduce false positives in the validation

process
• Achieved a remarkable 90% accuracy in photo validation, significantly improving efficiency and Automation

• Helped to reduce CS head count by 90% dedicated to photo validation, helping cost cutting to the firm.

Profile Segmentation | Excel, Vertica, Hive

• Created segment classification to identify high converting segments.
• Able to identify 30% of profiles making up 80% of total revenue
• The project helped the business in many ways identifying potential payment leads, promoting better packages
and decreasing the non-converting calling costs
Fraud Detection | Python, XGBoost, Django, SQL
• Lead the team to detect frauds by way of analyzing the customer input and activity.
• Developed an XGBoost based model using profile demographics and day to day activity of the user like interest
sent, frequency of interest sent, login count, login interval to analyze and classify based on the patterns between
a genuine and fraudulent user
• Developed a script to identify photo based fraudulent profiles and their duplicate profiles.
• Also studied and developed rules-based conditions to check fraud patterns.
• The project also helped the company to implement various Fare Usage Rules to mitigate fraudulent users.
Recommendation Engine | Python, Django
• Recommend the Best matches to the Customers in the Matchmaking Platform.
• Delivered a recommendation model to recommend profiles based on user data and preferences for assisted
services collaborating with product team.
• Helped Increase the Customer stickiness in the App by Average 2 minutes per day.
• Increased the Cohort Conversion by 1% translating to 10 million additional revenue every month and able to sell
High ARPU.
Bio Generation | Python, llama2 7B, PEFT, PyTorch, QLoRA, LLM, NLP
• Developed a system that automatically generates user-specific bios.

• Fine-Tuned LLaMA 2 7B large language model, to ensure high-quality and contextually relevant bio generation using

QLora technique.
Profile Validation | Python, XGBoost, AWS Bedrock, NLP
• Part of the team to develop and implement an Automated profile Validation API using a combination of open-source

Transformer and in-house-developed XGBoost models.

• Implemented ID verification to capture the name loaded on the portal with the input given in the registration page for

Aadhar, PAN and DL by using Claude 3.5 Sonnet model.

• Leveraged AWS Bedrock and Lambda services to implement and deploy ID verification model providing serverless and

cost-efficient solution
• The project helped in reducing the required manpower by 80% and help save cost to the company.
 Vernacular Transliteration | Python, Amazon EC2, Django, Transformer, NLP
• Leveraged an open-source transformer model to transliterate text from various vernacular languages across India

into English.
• Deployed API over Amazon EC2 instance for production, Used Django framework to build the API.

• Boosted Vernacular profile validation to 78%.

Vehicle image synthesis | Python, Tensorflow, Cuda, GAN, AWS EC2, DeepLabV3
• Finetuned Nvidia StyleGAN based GAN model to generate vehicle images using millions of input image data
• Developed and deployed a custom image background removal solution by finetuning DeepLabV3 image
segmentation model which serves to create clean car images corpus for GAN training time optimization
• Built a pipeline for this project using Remote SQL, SCP and Python

Education
Indian Institute of Technology Kharagpur (IITK) Kharagpur, IN
M.Tech Jun. 2016 – May 2018
Jawaharlal Nehru Technological University Vizianagaram (JNTUV) Vizianagaram, IN
B.Tech Jun. 2012 – May 2016

Certifications
Natural Language processing with sequence models – Deeplearning.ai
Natural Language processing with attention models – Deeplearning.ai