Data Science Intern

Technical duties

• Design and execute exploratory data analysis across diverse financial datasets using Python libraries such as pandas and NumPy.
• Develop conceptual and predictive models, statistical analyses, and optimization routines leveraging scikit-learn, PyTorch, and TensorFlow.
• Implement natural language processing workflows and LLM integrations using LangChain, Hugging Face, and Unsloth.
• Build, clean, and normalize ETL data pipelines to ingest diverse structured and unstructured data sources safely and efficiently.
• Perform rigorous hypothesis testing and experiment iteration to validate machine learning model outputs.

Project responsibilities

• Developing client recommendation engines and anticipating client needs using historical transaction and interaction datasets.
• Building automated market commentary generation tools and natural language processing pipelines using modern transformer architectures.
• Designing optimal risk management models and regime disentanglement frameworks for financial mixture of experts and interest rate swap portfolios.

A typical day (Karyvio’s outline for this kind of role — not an employer commitment)

• Attend daily lab stand-ups to align on project milestones, data experimentation results, and blocking issues.
• Write and optimize Python code for data ingestion, cleaning, normalization, and exploratory data analysis using pandas, NumPy, and scikit-learn.
• Design, train, and evaluate machine learning, deep learning, or NLP models (leveraging PyTorch, Hugging Face, langchain, or unsloth).
• Run hypothesis tests and structured experiments, documenting performance metrics and failure modes.
• Review model outputs with mentors and business stakeholders, gathering user feedback to iterate on predictive features.

Skills that matter, and why

• Python

  Serves as the core programming language for data manipulation, building predictive models, and deploying machine learning workflows.

  How to learn it: Build end-to-end data pipelines, practice coding problems, and learn efficient data structures using NumPy and Pandas.

• Statistics

  Provides the mathematical foundation necessary for hypothesis testing, probability distributions, inference, and understanding model behavior.

  How to learn it: Study probability theory, linear algebra, and statistical inference through online university courses and applied statistical problem solving.

• Machine Learning

  Enables the construction of classification, prediction, and clustering models to extract valuable insights from diverse datasets.

  How to learn it: Implement standard algorithms using Scikit-Learn and participate in platform challenges like Kaggle.

• PyTorch

  A primary deep learning framework required for building complex neural network architectures and generative models.

  How to learn it: Follow official tutorials to build simple feedforward networks, convolutional networks, and custom training loops.

Preparation plan

1. Strengthen core programming skills in Python, focusing on writing clean, concise, and production-ready code.
2. Master data manipulation and analysis workflows using Pandas and NumPy.
3. Brush up on statistical concepts, hypothesis testing, and exploratory data analysis techniques.
4. Build practical familiarity with deep learning frameworks and transformer libraries such as PyTorch, Hugging Face, and LangChain.
5. Review common machine learning procedures including classification, prediction, clustering, and neural network architectures.

Project ideas to show you can do the job

Interview guidance

• Revise fundamental concepts in probability theory, linear algebra, and statistical inference.
• Practice implementing machine learning algorithms from scratch or using Scikit-Learn with a focus on understanding why models work.
• Be prepared to discuss your past academic projects, Kaggle notebooks, or contributions to open-source data science communities.
• Demonstrate your coding proficiency in Python and familiarity with deep learning frameworks like PyTorch or TensorFlow.