Data Scientist Machine Learning

Sri Harshitha Yannam

San Jose, CA — ******************@*****.*** — 414-***-**** — LinkedIn — GitHub Data scientist and AI engineer with expertise in Artificial Intelligence, Machine Learning, and Software development. Proficient in Python and advanced AI, ML frameworks. I excel in dynamic environments by leveraging my technical and problem-solving skills to develop innovative AI solutions.

SKILLS

Programming Languages: Python, Java, C, C++

Web Development & Automation: HTML, CSS, JavaScript, Selenium, BeautifulSoup Database Systems: MySQL, PostgreSQL

Artificial Intelligence & Machine Learning Techniques: Supervised, Unsupervised, Reinforcement, ANNs, DNNs, CNNs, RNNS, Transformers, BERT, RAG, BART, CNNs, RNNs, LSTMs, GRUs, GANs AI Frameworks & Tools: OpenCV, LLaMA Index, ChromaDB, FAISS, LangChain, Diffusers, Flux, OpenAI, Transformers Libraries: NumPy, Pandas, Scikit-Learn, TensorFlow, Keras, PyTorch, Flask, SQLAlchemy, Matplotlib Security & Networking: Vulnerability Assessment, Jailbreaking Techniques, SSH EXPERIENCE

AI Engineer, HydroX AI, USA Febrauary 2024 – Present HydroX AI is focused on securing LLMs by providing a platform to evaluate, mitigate, & monitor potential risks related to safety, privacy, integrity, and security within AI systems.

Project - AI-Powered Memory Assistant Chatbot

Developed an AI-powered memory assistant to capture and store user interactions and usage patterns.

Handled advanced NLP architectures (BERT, RAG, BART) and leveraged LLMs for efficient information retrieval & session management.

Enhanced chatbot capabilities using Transformers, LangChain, LLaMA Index, & vector stores (ChromaDB, FAISS) for human-like responses.

Fine-Tuned models, incorporated Incontext learning on models (like GPT, Gemma), Prompt Engineering Techniques to handle the LLMs.

Automated data ingestion with Selenium and BeautifulSoup, integrating multiple AI models to significantly improve user experience. Research & Implementation - LLM Vulnerability Assessment and Attack Methods

Created a novel attack method combining persuasive tactics and cipher-based techniques (Caesar, ASCII, Morse, UTF) to generate exploitable image vulnerabilities from textual inputs.

Utilized Diffusers, Flux, GPT-4, and Llava to assess PCipher's effectiveness, ensuring cost-effective and scalable evaluations with both open-source and proprietary models like DALLE-3.

Conducted research on adversarial attack techniques targeting LLMs, focusing on their evolution and mitigation strategies in generative AI.

Explored methodologies including Backdoor Attacks, Adversarial Attacks, Jailbreaks, Prompt Injection, TAP, PAP, and Artprompt.

Identified vulnerabilities in AI bots (Flying Homes, Booking.com, TikTok’s Duobao) through diverse attack strategy experiments. Data Scientist, Mpowered Health, USA April 2023 - January 2024 Mpowered Health is a healthcare platform that connects consumers and healthcare organizations to improve healthcare. Project - AI-powered Chatbot for Customer Support

Developed a Rasa chatbot, increasing response precision by 25% and achieving an 85% success rate with human-like responses under 5 sec.

Enhanced chatbot functionality by annotating data and integrating ChatGPT APIs, improving range and responsiveness.

Built data pipelines to integrate chatbot data with Google Sheets, reducing response times by 30% and boosting efficiency.

Managed the entire SDLC for the chatbot project, overseeing requirements, deployment, and maintenance. Web Analyst and Data Scientist Intern, Nerv+ Company, USA August 2020 - December 2020 Nerv+ creates scientifically formulated over-the-counter products to alleviate stress and enhance focus, providing effective and accessible mental health solutions through grocery and convenience store channels. Project - Website Tracking using Google Analytics

Showcased Google Analytics applications for user interaction analysis.

Configured Google Analytics to monitor and report user interactions with 95% accuracy. Project: NELLY Conversational AI Platform

Reserached for developing an AI-powered platform for mental health support through safe, natural conversations.

Utilized a 3-billion parameter RNN, BERT-based retrieval, and custom ML models to detect emotions from user messages.

Categorized user emotions to enhance the bot’s ability to recognize and respond supportively. Data Scientist Intern, Techimax IT Services Pvt Ltd, India March 2020 - July 2020 Techimax is an IT services and product-engineering company specialized in building Hybrid cross-platform apps, using cutting-edge and emerging technologies.

Project - The Data Ninja - AI/ML Application

Improved accuracy through rigorous training and robust data pipeline implementation.

Applied SDLC principles and utilized techniques like SVM, Regression, Classification, and Clustering to develop, test, and deploy models, boosting system performance.

Employed Reinforcement Learning for performance optimization and created visualizations to simplify complex data insights. EDUCATION

Master’s in Computer Science, University of Wisconsin Milwaukee, USA August 2022 – May 2024

Classes taken: Advanced Machine Learning Natural Language Processing Networks and Database Security Algorithm Design Analysis.

Bachelor of Computer Science Engineering, Vaagdevi Engineering College – JNTUH, India August 2018 – July 2022

Classes taken: Machine Learning Data Structures Design Analysis and Algorithms Computer Networks Operating Systems Java C++

C PHP Software Engineering.

ACADEMIC PROJECTS

UWM CSI November 2023 - December 2023

Project - Chatbot Development for CSI Advanced Manufacturing testbed

Collaborated on developing an educational chatbot for a CSI Advanced Manufacturing testbed, utilizing Python, Azure, OpenAI’s GPT-3.5 APIs, Jupyter Notebooks, and Rockwell Automation manuals.

The project aimed to go beyond simple query responses, focusing on making complex manufacturing processes easy to understand. Vaagdevi Engineering College April 2021 - June 2021 Project - E-Waste Management/Motherboard Prediction

Implemented CNNs to achieve 82% accuracy in identifying damaged motherboard components.

Addressed electronic waste reduction by repurposing functional components and recycling non-functional ones. SELF PROJECTS

Skin Disease Identification Using Image Analysis

Created a deep learning system for identifying skin diseases from images, achieving 94% accuracy.

Implemented data preprocessing pipelines and annotated datasets trained with custom CNN architecture over 10,000 images by labeling and annotating images to train the model, enhance the model accuracy. Deep Learning Techniques for Breast Cancer Risk Prediction using IBM Cloud

Developed and deployed CNNs using TensorFlow and Keras for mammogram image classification, achieving 93% accuracy.

Utilized IBM Cloud for distributed, optimizing processing efficiency and scalability. Wounds Treatment Using CNN

Developed a CNN-based system that accurately classifies wound types and recommends appropriate treatments, enhancing the precision of medical interventions.

Designed image recognition and automated hospital recommendation, improving patient outcomes and facilitating timely medical assistance.

Predicting Employee Attrition Using Random Forest

Conducted feature engineering and data preprocessing and feature engineering for predictive modeling of employee attrition.

Developed a Random Forest model with 87% accuracy, providing actionable insights to improve employee retention strategies. CERTIFICATIONS & AWARDS

• CISCO CLA (Programming Essentials in C), March 2019

• IEEEXtreme 13.0 Programming Competition, October 2019

• Internal Hackathon for Smart India Hackathon 2020 (Cleared first level), February 2020

• IBM Badges and Certificates from Cognitiveclass.ai in Data Science, Data Visualization, Data Analysis, and Big Data, September 2020

• “Artificial Intelligence” program powered by 1Stop, IIT Kanpur, Aug 2021

• CSI Hackathon at UW-Milwaukee (3rd Prize), December 2023

Contact this candidate