Anushka Agarwal

Anything

About

Passionate machine learning engineer with a strong background in Computer Vision, AI and Software Development. With experience across startups, research and industry, I thrive at the intersection of technology and real-world impact. Masters in Computer Science graduate (Machine Learning track) from Columbia University, New York.

Machine Learning Engineer & Software Engineer

  • Birthday: 24 January 2002
  • Website: anushka24agarwal.github.io
  • Phone: +1 646 210 4265
  • City: New York, USA
  • Age: 23
  • Degree: Master's
  • Education: Columbia University
  • Email: agarwal.anushka@columbia.edu

I have over two years of experience building software and ML systems, spanning Software Engineering roles at startups and industry, along with research experience as a Machine Learning Researcher at the Columbia Climate School.

Winner of UNESCO India Africa International Hackathon 2022

Winner of Smart India Hackathon 2022 Software Edition

Research Paper published in IEEE Xplore [Link]

Skills

Programming Languages: C, C++, Python, Java, SQL, MATLAB, HTML, Bootstrap, React, JavaScript, TypeScript, LINUX

Cloud & DevOps: AWS (EC2, S3), GCP, Docker, Kubernetes, BigQuery, REST API Design, GitHub Actions, CI/CD

ML Libraries: TensorFlow, PyTorch, Keras, Scikit-learn, OpenCV, Pandas, NumPy, Matplotlib, NLTK, Librosa

Frameworks: Flask, Django, FastAPI, Docker, Kubernetes, Pydantic, PyTest, PostgreSQL

Application Tools: GitHub, Google Colab, Jupyter Notebook, Copilot, Cursor, Claude

Resume

Education

M.S. Computer Science (Machine Learning Track)

Aug 2024 - Dec 2025

Columbia University, New York

Coursework: Applied Machine Learning, Computer Vision, Deep Learning, Artificial Intelligence, Advance Spoken Language, Algorithms, Databases

Teaching Assistant: Advance Topics in Deep Learning

Activity: DSI Scholar Spring 2025, Columbia Build Lab Engineer Fall 2024

B.E. Computer Science and Engineering

Dec 2020 - June 2024

R.V. College of Engineering, Bangalore, India

Coursework: AI and ML, Artificial Neural Networks, Object Oriented Programming, Advanced Algorithms, Data Structures and its Applications, Operating Systems

Activity: Student Placement Coordinator, Operations Manager at Frequency Club, Footprints Dance Club

Professional Experience

Software Engineer Intern

Sep 2025 - Dec 2025

GreenPortfolio, New York, NY

  • Automated ETL pipeline consolidating 1000+ financial records from APIs and internal feeds reducing manual entry by 95%.
  • Architected climate-score ingestion system that automated JSON exports from Bubble.io, storing 100+ records in cloud storage.
  • Designed normalized BigQuery data model for client-advisor matching accelerating match lookups by 70%.
  • Implemented unit tests and a CI/CD pipeline using GitHub Actions, increasing code reliability across production releases.

Machine Learning Researcher

Jan 2025 - Aug 2025

Columbia Climate School, New York, NY

  • Apply OpenCV-based image pre-processing to enhance segmentation across 1000+ marine microscopy images.
  • Build a transfer learning architecture via ResNet-50 leveraging TensorFlow for feature extraction.
  • Optimized K-Medoids clustering algorithm to identify 20+ distinct marine species with 85% purity score.
  • Develop production-ready FastAPI web service to serve ML pipeline through RESTful API endpoints.
  • Implement Docker-containerized app on GCP, reducing manual analysis workflow by 80% for end-users.

Software Engineer Intern

Sep 2024 - Dec 2024

Threepio, Columbia Build Lab, New York, NY

  • Launched a full-stack Django-React web app for subtitle generation, reducing processing time by 75%.
  • Engineered a Python based microservice architecture handling 100+ concurrent transcription API requests.
  • Programmed JWT authentication and RBAC system in Django REST Framework enabling secure multiple-client access.
  • Deployed services on AWS EC2 and configured CloudWatch dashboards and alarms ensuring continuous uptime.

Software Engineer Intern

Jan 2024 - Jun 2024

Intel, Bangalore, India

  • Built Docker containers and deployed test environments with Kubernetes, speeding regression testing cycles by 45%.
  • Integrated automated test execution into CI/CD pipelines utilizing PyTest and Selenium, reducing manual QA time by 70%.
  • Enhanced Python APIs for internal Calculation Workbench tool for enterprise workflows to support 100K+ invoice records.
  • Delivered 3 production-ready features in structured Agile sprints while collaborating under 2 senior engineers on code reviews.

Projects

ColumbiaStream

a cloud-based academic video platform designed to allow students and faculty within the Columbia community to securely upload, store, and access recorded course content. Inspired by existing video platforms such as YouTube, ColumbiaStream focuses specifically on academic use cases, providing role-based access and structured organization of lecture materials.

Driver Behavior Detection Model

Developed a driver behavior classification system using a customized DenseNet121 architecture, incorporating GlobalAveragePooling and Dropout layers for improved generalization. Achieved 98.42% test accuracy across six behavior categories through TensorFlow-based model training and OpenCV-powered image preprocessing and augmentation.

AI Fitness Planner

A GenAI-powered app that uses multi-agent architecture and open-source LLMs (Mistral via Ollama) to generate fully personalized diet and workout plans. Built with Python and Streamlit, it supports structured JSON outputs, Q&A follow-up and PDF export

URL Shortener Application

A full-stack URL shortening service built with FastAPI (backend), React (frontend) and PostgreSQL for storage. Fully containerized using Docker and Docker Compose. Generates short, shareable URLs, redirect users to original links with a modern React UI with Axios integration

Linguistic Bridge

Trained custom CNN model for ancient script recognition on 200+ unique character classes with data augmentation practices. Implemented thresholding and filtering techniques to pre-process 100+ scanned historical images leveraging OpenCV library. Tuned model hyperparameters through cross-validation, boosting classification accuracy from 70% to 85%.

Service Booking Platform

Developed intuitive marketplace platform with React frontend, implementing dynamic form validation and state management. Integrated RESTful Flask APIs for core platform operations, improving user workflow efficiency by 30%. Designed a normalized PostgreSQL schema and optimized SQLAlchemy queries, reducing join overhead by 40%. Hosted the application on GCP Compute Engine and Cloud SQL for scalable and reliable performance.

Contact

Location

Manhattan, New York, NY 535022

Call

+1 646 210 4265

Email

anushka24agarwal@gmail.com