Machine Learning Engineer

Utkarsh Tamrakar

813-***-**** ************@*****.*** https://www.linkedin.com/in/utkarshtamrakar https://github.com/utkarshb95 PROFESSIONAL SUMMARY

Machine Learning Engineer with a master’s degree in computer science and 3 years of hands-on experience in developing and deploying ML solutions. Expertise in designing and optimizing ETL and ML pipelines, fine-tuning large language models, and implementing production-ready models. Skilled in leveraging advanced statistical analysis and cloud infrastructure to drive innovation and deliver impactful results. Strong problem-solving abilities, collaborative team player, and dedicated to continuous learning and improvement. TECHNICAL SKILLS

Programming Languages: Python, R, C++, JavaScript, Java, SQL, Scala, Linux, REST API, Django, Flask ML & AI: TensorFlow, PyTorch, Scikit-Learn, Keras, XGBoost, NumPy, Pandas, Prophet, ARIMA, NLP, NLTK, OpenCV, Ensemble Methods, CNN, LSTM, Vision Transformer, OpenAI, Generative AI, LLM, RAG, RLHF Data engineering: Azure Data Factory, Synapse, Apache Spark, Hadoop, Kafka, Airflow, ETL, Redshift, Snowflake, BigQuery, PostgreSQL, MongoDB, Cassandra, HDFS, Azure Data Lake, MySQL Data Analysis & Visualization: Excel, Tableau, Power BI, Plotly, Seaborn, Matplotlib Cloud Tools: AWS, Google Cloud, Azure, Azure ML Studio, Databricks, Heroku, Azure DevOps Development Tools: Git, Docker, Kubernetes, Jenkins, VS Code, Jupyter, CI/CD, JIRA, MLFlow, IoT Edge PROFESSIONAL EXPERIENCE

Machine Learning Engineer, Gentherm (August 2022 - January 2024)

● Collaborated with stakeholders to develop ML solutions for automotive thermal technology, achieving 75% energy reduction and improving user comfort with regression algorithms and reinforcement learning.

● Engineered ETL & ML pipelines for IoT devices, reducing processing time by 50% and enhancing data management with PostgreSQL, Azure Data Lake, and Docker.

● Implemented CI/CD practices and optimized pipelines using Spark, Azure Data Factory, and Azure ML Studio, improving development efficiency.

● Deployed and monitored models in production for real-time insights and optimization, applying advanced statistical and time series analysis to reduce user comfort attainment time by 25%.

● Designed secure cloud infrastructure with Terraform and automated testing and deployment with GitHub CI/CD.

● Fine-tuned LLM for domain-specific applications, enhancing content generation and information retrieval with RAG, LangChain, nltk, BeautifulSoup, and Flask.

● Developed an interactive voice-controlled system with GPT-4 LLM, decreasing manual climate adjustments by 30% and improving response time by 25%.

Machine Learning Analyst, X2 Analytics (January 2022 - May 2022)

● Formulated NLP and ML techniques to re-engineer 50% of SQL database entries for anomaly detection.

● Led data collection and web scraping efforts, providing NLP-driven analytics for B2B clients.

● Enhanced sentiment analysis by 20% using advanced NLP techniques like NLTK and transformer models.

● Implemented NLP-based anomaly detection methods, including isolation forests and one-class SVM, reducing false positives by 30% and enhancing data quality assurance. Machine Learning Research Assistant, USF (August 2021 - May 2022)

● Tailored deep learning models including VAEs, GANs, and vision transformers for multi-object grasping with robotic hand and tactile sensors, achieving 80% accuracy through simulation-to-real fine tuning.

● Improved neural network accuracy by 20% through feature engineering and fine tuning on large datasets (>100k) from simulation systems, optimizing for robotic manipulation tasks. 2

● Experimented with diverse deep learning techniques, reducing computation time by 30% by addressing challenges like imbalanced datasets, noise removal, and dimensionality reduction.

● Leveraged distributed computing on an HPC system for parallelized simulations and model training across GPU nodes, resulting in a 70% reduction in training time. Data Engineer, 3G Solutions (July 2018 - December 2019)

● Engineered custom OCR software with OpenCV for automating marksheets evaluation, reducing manual workload by 40% and improving assessment efficiency.

● Utilized NumPy and Scikit-learn libraries to develop machine learning algorithms for data processing, contributing to a 3x increase in company revenue.

● Reduced downtime by 25% with predictive maintenance ML modeling including time series analysis and ensemble learning.

● Spearheaded CNN-based quality control initiatives leveraging OpenCV, resulting in a 40% decrease in defects. EDUCATION

M.S., Computer Science University of South Florida May 2022 (GPA: 3.95/4.00)

● Thesis Topic: Prediction of the number of objects in a robotic grasp B.Tech., Computer Science Jawaharlal Nehru Technological University Hyderabad, India May 2018 PROJECT EXPERIENCE

Reinforcement Learning

● Developed Q-learning neural network and a Double DQN agent beating Atari Pong after 56 hours of training using OpenAI gym with PyTorch.

Sequence Prediction

● Constructed LSTM and GRU models to estimate changes in pouring behavior with Keras and TensorFlow, achieving a 0.01280 RMSE on the test dataset.

Image classifier

● Devised a convolutional neural net for food item state classification with a 77% accuracy rate using PyTorch, involving ETL pipeline, image preprocessing, and performance optimization for ingestion and inference.

Contact this candidate