Data Scientist Machine Learning

Sandeep T Shelvan

***************@***.*** +1-412-***-**** https://www.linkedin.com/in/sandeep-t-shelvan/ EDUCATION

Carnegie Mellon University Pittsburgh, USA

Master of Science in Data Analytics for Science

Indian Institute of Technology, Guwahati Assam, India Master of Science in Physics with Specialization in Particle Physics WORK EXPERIENCES

● LvlUp Ventures Remote, USA

Senior Venture Fellow October 2024 - Present

● Market Research & Due Diligence – Conducted in-depth research on early-stage startups across various industries, analyzing market trends, competitive landscapes, and financial viability.

● Developed strong analytical skills in assessing business models and investment opportunities.

● Deal Sourcing & Pipeline Management – Identified and evaluated high-potential 15+ startups Financial & Business Analysis – Examined revenue streams, and key performance indicators (KPIs) to provide data-driven recommendations.

● Data Scientist Intern Pittsburgh, PA

Ansys Inc January 2024 - May 2024

● Collaborated with four students under the guidance of Ansys Data Scientists, using Python on AWS EC2 instances for Exploratory Data Analysis (EDA) and data cleaning on hardware simulation configuration data, to predict time and memory.

● Used Matplotlib and Seaborn for visualization, and Amazon RDS (SQL) for querying structured data.Developed and refined 45 machine learning models using scikit-learn and deep learning using TensorFlow training and deploying them via Amazon SageMaker.

● Utilized AWS S3 to store and manage synthetic datasets generated with Synthcity, expanding the dataset by 20% and improving model robustness.

● Achieved 98.7% accuracy (RMSE) in predicting computation times and 97.9% (RMSE) in forecasting memory usage by using XGBoost. Also, Random Forest gave the second best result.

● Space to Space Inc: A Deep-tech start-up Pittsburgh, PA CEO & Co-Founder May 2023 - Jan 2025

● As an international first gen student I was able to bring former NASA CFO, Ex-Senior executive from Northrop Grumman and former Deputy Inspector general of the Department of Air Force Board of directors and advisors.

● A non-rocket launch technology start-up that aims to cut launch costs to 1 by 15 the price of SpaceX. It was the first-ever kinetic launch system spin-off from a US university.

● Able to cold mail more than 200 VCs and Angels then secured 20+ more calls with them, able to chat with Vinod Khosla and a few other billionaires and got great feedback.

● Made executive summary, pitch deck, financial modelling, financial planning, and business plan for the company.

● Selected into the NVIDIA Inception Program and recognised as a start-up by the government of India and the state of Kerala.

● Able to get LOIs worth $300k. Got into the final interview round in Techstars NYC, Antler Canada, EWOR like accelerators.

● Joined Swartz Center and Project Olympus Incubator at Carnegie Mellon as well as University of Pittsburgh Incubation Center. PROJECT EXPERIENCES

● Teaching BERT Model to do addition operations January 2024 - May 2024

● Transformer models are not great at Maths. Tried to do addition using the Chain of thoughts method on the BERT model.

● Build an additional table and train on Bert model. Achieved 0.93 of F1 score for two digit addition.

● Arxiv title prediction January 2024 - May 2024

● Engineered LLMs to predict ArXiv titles from abstracts by training on ArXiv metadata using Llama3, GPT, Cohere AI and bert.

● ROUGE and BLEU scores were used for evaluation; the bert model performed poorly, while Llama-3 achieved a BLEU of 0.9.

● Protein family classifications January 2024 - May 2024

● Classified protein by fine tuning Protein Bert model from hugging face by tokenizing amino acids. This multi-class classification attained 100% accuracy.

● Large Scale Parallel Computing for 2D heat distribution August 2023 - November 2023

● Operated a supercomputer at the Pittsburgh Supercomputing Center, leveraging PySpark for parallel computing across multiple nodes to solve heat distribution in a 2D metal plate using the Finite Difference Method.

● Queried and analyzed large simulation datasets, optimizing data exploration and validation using Spark SQL.

● Optimal Zip code for drone stations August 2023 - November 2023

● Used MySQL queries on a large dataset for drone delivery to identify optimal zip codes where the number of deliveries is high and which fall within the drone's range, while also generating maximum profits. RESEARCH EXPERIENCES

● Master’s Thesis: Transverse Momentum Distribution of Color Singlet Final State at Large Hadron Collider.

(High Energy Physics, Phen) August 2021- May 2022

● Using monte-carlo simulations on super computer clusters to simulate 300GB of subatomic particle collision data.

● Used Python for Exploratory Data Analysis (EDA) and machine learning techniques to clean and preprocess QED data.

● Utilized SQL to query and organize large datasets generated during simulations for efficient analysis.

● Created visualizations using Tableau to communicate findings effectively, and utilized LaTeX and PowerPoint for presentations.

● Bachelor’s Thesis: Green Synthesis of Silver Nanoparticles and Characterization. January 2020 - May 2020

● Worked alongside three other students to collect data from UV Spectroscopy. Cleaned and transformed the data using numpy and pandas.

● Conducted data analysis and visualization on silver nanoparticle synthesis using, Excel, and ggplot2, matplotlib interpreting X-ray diffraction patterns to confirm nanoparticle composition. Presented and submitted the thesis using word and powerpoint presentation.

SKILLS

Programming Languages/Libraries: SQL, Python, R, TensorFlow, NumPy, Pandas, Apache Spark, Scikit-learn, Keras, PyTorch, MPI programming, seaborn, Matplotlib, SciPy, NLTK, SpaCy, PySpark, SparkSQL, SparkML, Tableau, ggplot2. Software Used: Matlab, Mathematica, Microsoft Excel, Microsoft Word, Microsoft PowerPoint, Microsoft Office, Git, R Studio, Power BI. Generative AI Skills: Bert Transformers, Hugging Face, Langchain, Fine tuning of LLMs,Chain of Thoughts, Chain-of-Trees, Ollama, CrewAI. RELEVANT COURSEWORK

Machine Learning, Big Data, Data Structures and Algorithms, Advanced Linear Algebra and Tensor Analysis, Advanced Calculus, Deep Learning and Neural Networks, Prompt Engineering, Big Data, MLops.

Contact this candidate