Machine Learning Research Assistant

Srinivas Pendotagaya

Denton, TX, *****, +1-682-***-**** **************.***@*****.***, LinkedIn: www.linkedin.com/in/pendotagaya-srinivas

Dedicated AI Programmer with 8+ years of experience in Python, C/C++, SQL, MATLAB, and Arduino programming. Skilled in machine learning frameworks such as TensorFlow, PyTorch, and Scikit Learn, with expertise in RNNs, CNNs, Transformers and LSTMs. Proficient in software tools like PyCharm, Jupyter Notebook, and Visual Studio, and experienced in hardware platforms like Arduino and Raspberry Pi. Familiar with cloud platforms including Microsoft Azure AI Fundamentals and AWS, as well as containerization tools like Kubernetes and big data frameworks such as Apache Spark

Work Experiences

UNIVERSITY OF NORTH TEXAS

Research Assistant and Instructional Assistant

CoVIS Lab, UNT (Sep. 2022 - current)

Lead Developer and Designer

I'm involved in a NASA-funded project to improve water-use data collection for crop irrigation, covering 20% of US cultivated land but using 42% of freshwater withdrawals. Using a dataset of annotated agriculture tiles (5000x5000), our classification model achieved 98% accuracy, surpassing state-of-the-art models in irrigation practice type classification and image segmentation.

TECH iNSPiRON

Lead NLP Engineer (July 2016 – July 2022)

Utilized diverse techniques including SVM, Logistic Regression, LSTM, and CNN, optimizing them for text classification. Leveraged tools like scikit-Learn, Keras, and TensorFlow in CPU/GPU environments, achieving state-of-the-art performance. Utilized advanced techniques such as Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), and Transformer-based architectures to achieve this goal. Implemented reproducible models through Docker, deploying them on AWS ECR Repositories/S3 artifacts using Jenkins CI/CD orchestration. Automated data pipelines, integrating NoSQL databases, APIs, and ML engines for seamless data science workflows. Applied transfer learning methods with pre-trained word embeddings, Large Language Models (LLMs) such as Flan-T5 and ChatGPT for tasks like text similarity and natural language understanding. Led the development of sentiment analysis systems and chatbots for real-time customer feedback analysis, integrating sentiment analysis into chatbots for personalized responses based on customer sentiment. Utilized LLMs like Flan-T5 and ChatGPT for nuanced understanding and context-aware interactions, significantly enhancing chatbot capabilities. Deployment on cloud platforms reduced response time by 30%, garnering positive feedback from stakeholders for improved customer satisfaction.

EMPORIS

Application Developer (Nov 2014 – June2016)

I've worked extensively with Arduino, Raspberry Pi, and IoT technologies. Collaborating closely with the development team, I've contributed to product development, application support plans, and prototype programs. Additionally, I've partnered with business unit team members to design new application systems, focusing on enhancing client requirements for mobile computing capabilities. My role also involves collaborating with software development and testing team members to design and implement robust solutions that meet client requirements for functionality, scalability, and performance.

Projects

Siamese Net Mapping of Irrigation Practices in Satellite Images

CoVIS Lab, UNT(June 2023 - Jan 2024)

Lead Developer and Designer

Pioneered Siamese Network research for agricultural field classification. Optimized a modified ResNet50 backbone for binary and three-class tasks. Introduced an adaptive contrastive loss function, enhancing adaptability and interpretability. Applied findings in precision agriculture, emphasizing real-time monitoring and agile responses. Made impactful contributions to model architecture and precision agriculture applications.

Irrigation Practice Mapping from Satellite Images Using Deep Convolutional Networks

CoVIS Lab, UNT (Sept. 2022 - May. 2023)

Lead Developer and Designer

Implemented an innovative segmentation approach using field segment super pixels, highlighting the effectiveness of a fusion model with ResNet-50 and VGG-16. Employed Felzenszwalb's super pixel method and SamGeo for geospatial field segmentation. Explored the impact of input patch size on segmentation performance. Trained and compared ResNet-50, VGG-16, Inception-V3 models for image classification, with ResNet-50 excelling. Summarized segmentation performance metrics, focusing on accuracy, balanced error rate (BER), and mean Intersection over Union (mIOU).

Landscape Generation with GANs

Developed a generative adversarial network (GAN) to create realistic images of imaginary landscapes for virtual reality environments. Utilized deep learning techniques to train the model on a dataset of diverse landscapes and fine-tuned it to generate novel and visually appealing scenes, enhancing the immersive experience for users.

Sentiment Analysis with BERT

Implemented a sentiment analysis system using HuggingFace's pre-trained BERT model to classify customer reviews in an e-commerce platform. Integrated HuggingFace's Transformers library to fine-tune the model on a labeled dataset and deployed it as a service, providing real-time sentiment analysis for incoming reviews, thereby enabling the company to better understand customer feedback and improve product offerings.

Bias Mitigation in Loan Approval System

Collaborated on a project to assess and mitigate biases in a loan approval system using the Fair learn framework. Analyzed historical loan data to identify disparities in approval rates across demographic groups and employed Fair learns fairness metrics to quantify and address these biases. Implemented fairness-aware machine learning algorithms and evaluated their performance to ensure equitable decision-making in the lending process.

Publications

An Accelerometer Based Digital Pen with A Trajectory Recognition Algorithm For Handwritten Digit And Gesture Recognition

Wireless Intelligent Accurate Bridge Deck Crack Inspection and Mapping

Areas of Expertise

Machine Learning (advanced), Deep Learning (advanced), Image Processing (advanced), Natural Language Processing (NLP)(intermediate), Data Mining (advanced), Time-Series Forecasting (advanced), Lang chain(intermediate), Data visualization (advanced), Neural Networks (advanced), Information Retrieval and Extraction (advanced), Predictive Modeling (intermediate),LLM(intermediate), RAG(intermediate),RLHF(intermediate),Langchaine(intermediate).

Programming Languages

Python, C, C++, SQL, Arduino Programming, MATLAB, Embedded C,R

Software’s:

PyCharm, Jupyter Notebook, Visual Studio, code blocks, MATLAB, MySQL, Arduino, Raspberry pi

Tools and Devices:

Scikit Learn, TensorFlow, PyTorch, CNTK, Kera’s, Apache MX Net, Arduino,RNNs, Node-RED, Raspberry pi, Esp8266, Opencv, Kubernetes, Apache and Spark,CNN,LSTM

Cloud:

Microsoft Azure AI Fundamentals, AWS

Certifications

Internet of things (2019) -cognixia

Advanced Deep Learning and computer Vision (2021) -SimpliLearn

Machine Learning (2021) by Professor Andrew N.G. (Coursera)

Data science with Python (2021) -SimpliLearn

Deep Learning with Tensor Flow (2021) -IBM

Deep Learning (2021) -SimpliLearn

Al Capstone (2021) -SimpliLearn

Deep Learning with Kera’s and TensorFlow (2021) SimpliLearn

Machine Learning Advanced Certification Training (2021) SimpliLearn.

Natural Language Processing (2021) SimpliLearn

Natural Language Processing in TensorFlow (2024) Coursera

Generative AI with Large Language Models (2024) Coursera

Education

M.Sc. in Artificial Intelligence (GPA: 4.0/4.0), University of North Texas, Denton, TX, U.S. (Aug. 2022 – May. 2024)

M.Sc. in VLSI and Embedded Systems (GPA: 3.3/4.0), JNTUH, Hyderabad, Telangana, India (June 2012 – July 2014)

Bachelors in Electronics and Communication Engineering (GPA: 3.06/4.0), JNTUH, Hyderabad, Telangana, India (Aug. 2008 – March 2012)

Contact this candidate