# Anudha M. > Engineer / Machine Learning Research / Web Development / Leadership / Consulting Location: San Francisco Bay Area, United States Profile: https://flows.cv/anudha The number of things I don't know is perpetually increasing. https://scholar.google.it/citations?user=9joZ0ioAAAAJ&hl=en My wiki info: https://sites.google.com/view/anudha/about-me Interest: Signal Processing Natural Language Processing (NLP) Computer Vision Image analysis Electron Microscopy Statistical Mechanics Stochastic methods for any application Materials science and physical sciences. Machine translation. Also interested in the impact of arts - paintings, literature, cinema Work details: • Developed a web framework for non-developers to interact with model training • Evaluate: training time, inference time, model size Optimize model performance: confusion matrix, training loss Train Small Models from Scratch with Audio, Vision, and Text Data Training environment: Ubuntu, AWS instance. Maintain ssh session via tmux. • Data ingestion, processsing, model development using numpy, scikit-learn, tensorflow, pytorch, jax, pandas, numba, • Decrease size of model by pruning nodes and edges • Use above results to decide on training larger models and compute budget • Contribute to code repo via git Increase Accuracy of Machine Translation • Fine-tuned models on GCP and Azure. Tested open source models hosted on Hugging Face. • Data Preparation (Web scraping, alignment of multi-lingual sentence-pairs using LABSE) • Implement performance metrics (cross-lingual similarity, Blue score) • Get feedback on translation quality and usability from stakeholders • Iterate on the tool based on user requirements • LLMs and prompt engineering for translation Data graphing and visualization • Rendering large data sets with many dimensions has challenges in speed and interpretability Delivered solutions with python modules: plotly, seaborn Audio processing: • Calculate pitch of audios as a music learning aid. Connects python machine learning and artificial intelligence models to html frontend. • Configured apache2 web server • Strings together yt-dlp, ffmpeg for stripping audio from video, librosa for reading audio, and FFT (Fast Fourier Transform) analysis. • Librosa has a dependency on numba. Numba writes files to a dir where webserver doesn’t have write permission. Diagnosed problem because librosa functionality was working via terminal, but not via web server. Modified numba configuration to resolve. Tested open source codes for sound separation. ## Work Experience ### Engineer @ Samsung Research America (SRA) Jan 2025 – Present Development and test related to AI-Web, Browser, Chrome extensions, Chrome APIs, Natural Language Processing (NLP) ### AI Solutions Engineer @ Centific Jan 2025 – Present ### Machine Learning Engineer @ Caterpillar Inc. Jan 2023 – Jan 2023 Fine-tuned machine translation models hosted by Azure and GCP for European and Asian languages Developed multi-lingual datasets Evaluated semantic similarity models (LASER, LABSE, SONAR) to use for evaluating quality of translation when a human reference is not available. Identified unsolved problems in translations and future directions. Evaluated performance of open source AI models (inference time, accuracy, Blue Score) Used LLMs to clean training datasets Build semantic search / cross-lingual semantic search ### Artificial Intelligence Consultant @ Penn State University Jan 2022 – Jan 2022 Implemented AI Models and ML algorithms for different domains and use cases (Econ, Bio) The size/scale of datasets/models was restricted to academic research. Made a pipeline that scraped 25k urls based on user search criteria in 2 days on a node with 40 cores. Encoded the text. Fed the data into ML classifiers. Fitted the classifiers based on user-designed labels. Predicted labels on new text. Provided metrics on accuracy of model. Provided graphs that help grasp distribution of accurate predictions. ### Data Science Engineer @ Kaiser Permanente Jan 2020 – Jan 2020 Coded and demo’d prototype dashboard to take action based on leading indicators and refute hypotheses not supported by data. Leading indicators were identified using correlation and lag in internal/external time-series data. Automatic daily updates via cron job for continued tracking of real world data. Delivered code to be integrated into a more comprehensive dashboard. ### Senior Materials Science Engineer @ Naval Nuclear Laboratory (FMP) Jan 2015 – Jan 2019 ### Chemical Engineer @ GE Jan 2014 – Jan 2015 ## Education ### Doctor of Philosophy (Ph.D.) in Chemical Engineering University of Minnesota ### Bachelor's degree in Chemical Engineering University of Massachusetts Amherst ## Contact & Social - LinkedIn: https://linkedin.com/in/anudha-m-72199a18 --- Source: https://flows.cv/anudha JSON Resume: https://flows.cv/anudha/resume.json Last updated: 2026-04-11