Machine Learning Data Scientist

KUMAR KOUSHIK TELAPROLU

812-***-**** ****************@*****.*** https://www.linkedin.com/in/telaprolu-kumarkoushik/ https://github.com/WannaBeNeuralNetwork EDUCATION

Master of Science in Computer Science, Indiana University, Bloomington, CGPA: 3.8/4 Aug 2023 - May 2025 Deep learning Architecture and Hardware Acceleration, Data visualization, Advanced Machine learning in Computational linguistics, Applied Machine learning, Software Engineering, Applied Algorithms, System Protocols Service Assurance, Engineering Cloud Computing. Bachelor of Technology in Computer Science, Amrita Vishwa Vidyapeetham, Bengaluru, India, CGPA: 8.14/10 Jul 2018 – May 2022 TECHNICAL SKILLS

Programming Languages: Java, Python, Arduino, HTML, CSS, Java Script. AI & Data Science: ML and Deep learning techniques, Computer vision, Chatbots, Quantum NLP, LLMs, Recommendation system. Frameworks & Libraries: Langchain, Langgraph, CrewAI, Django, React.Js, Node.Js, Flask, FastAPI, ApacheSpark, CI/CD, TesnorFlow, PyTorch, Keras, PyG, Scikit-Learn, RDF, SparQL, Pandas, CUDA, BS4, Kafka, Reddis, Tableau, Gradio, BM25, Vector Search, ANN. Cloud/Databases: MySQL, PostgreSQL, MongoDB, AWS, GCP. Certifications: NVIDIA Associate GenAI LLMs, Professional Accelerated Data Scientist. PROFESSIONAL EXPERIENCE

ML SDE at Verizon Data Services, Hyderabad Scikit-learn, EDA, ML, Python, Java, Spring Boot, GCP, Mongo DB, 5G. Jul 2022 - Jul 2023

• Developed and optimized backend systems for managing outage ticket flows, deserializing and structuring customer complaint data for machine learning-driven analysis and routing.

• Designed data pipelines that enabled intelligent pre-resolution of 5G network outages, contributing to a 30% faster outage resolution and 20% higher automated ticket handling.

• Conducted root cause analysis on network outages, integrating new features and signals into machine learning models to improve predictive capabilities. Collaborated with cross-functional teams to ensure robust deployment pipelines using Jenkins, Kubernetes, and scalable backend architecture, laying the foundation for AI-driven troubleshooting initiatives. Student Intern at Verizon Data Services, Hyderabad React Js, Next Js, MongoDB. Feb 2022 - Jul 2022

• Developed modules of the application to manage various outages issues created by consumers and visualize the data in different statistical formats. Integrating different APIs.

Junior Data Scientist (Intern) at EXCELR SOLUTIONS, Bengaluru ML, Scikit-learn, pandas, NLTK, Data Mining May 2019 - Jul 2019

• Forecasted Client-Company sales by building and optimizing multiple machine learning models using Pandas, Scikit-Learn, and NumPy, optimizing techniques by hyperparameter tuning, batch normalization and learning rate scheduling. ACADEMIC PROJECTS

F1 Race Prediction Engine IUB Python, Scikit-learn, Fast API, Gradient Boosting, Feature Engineering.

• Built a Gradient Boosting ML model to predict 2025 F1 race outcomes using historical race and qualifying data; achieved a 3.22s MAE. Automated race-by-race updates through FastF1 API data ingestion and dynamic feature engineering. NutriSync – AI Powered Dietary recommendation app TensorFlow, Neo4j, LangChain, FAISS, Knowledge Graph RAG, Nvidia Nemo, OCR.

• Built a mobile AI assistant that scans food barcodes, cross-references user medical conditions, and generates personalized dietary warnings and safe alternatives in real-time.

• Designed a RAG pipeline integrating OCR/NER, FAISS semantic search, Neo4j food-disease graphs, and NVIDIA LLaMA 3.3 (via LangChain

+ NeMo) for structured health assessments. Visualized food risk profiles using React Native cart views and dynamic knowledge graphs, combining deep retrieval and graph reasoning for explainable dietary insights. AI Agent for operating backend systems with natural languages IUB Gradio, Slot, GPT-4, Llama 3.2, BERT, MongoDB.

• Developed Built an AI agent system that converts natural language inputs into database operations using BERT for intent detection and slot classification. Improved multilingual capability and contextual awareness via LLaMA 3.2 and GPT-4 integration. Evaluating Hallucinations on Hindi language IUB Term Paper, GPT-4, Llama 3.1, Hindi Hallucination QA Dataset.

• Developed Designed HHQA dataset for evaluating cultural hallucinations in Hindi using GPT-4 and LLaMA 3.1.

• Achieved 63.64% non-hallucination accuracy with GPT-4; introduced a benchmark for underrepresented languages in QA. TAGE – Transcribe, Analyze, Generate, Extract IUB GCP, BERT, BART, FASTAPI, Llama, MERN, OpenAI Whisper, Browser Extension.

• Built a real-time AI meeting assistant generate transcriptions, summaries, sentiment insights, and action items.

• Developed a Chrome extension and web app with React, Electron, and FastAPI to overlay context analytics for live meetings. Graduate Teaching Assistantship – Introduction to Python Programming, Big Data Analytics, Luddy School, IUB. Sep 2024 – May 2025 Graduate Research Assistantship IUB Sentimental analysis, NLP, Ontology, Knowledge graph RAG, Aug 2024 - Present

• Researching under Prof. Varun Sharma on sentiment analysis and data mining for the project "(Almost) 200 Years of News-Based Economic Sentiment," with IU's Big Red 200 supercomputer for large-scale text processing, neural network modeling, and economic insight extraction.

• Researching under Prof. Damir Cavar on a biomedical ontology-based knowledge graphRAG architecture with triplet-based RDF graphs, hypernym-driven ontology construction, chunk-specific question generation to improve retrieval relevance and semantic fidelity. Integrated query-to-question semantic alignment over SBERT, BioClinicalBERT, optimizing classical vector search in retrieval accuracy and hallucination resistance. Ontologies created from PubMed and NIH using SpaCy, UMLS-lite and dependency parsing to extract IS-A hierarchies and domain-specific relations

Contact this candidate