HELLO, I'M
Ananya Agarwal
Building reliable systems: full-stack apps, data pipelines, and evaluation-driven ML. Looking for Summer or Fall 2026 internships.
0+ Internships 0 Projects 0 Skills
Work Experience
Software Engineer
New York, USA
- Modernized a large-scale image intelligence platform from Angular 4 to Angular 20, migrating 31 components, 9 services, and 8 data models to a modular standalone architecture integrated with Firebase (Auth, Realtime DB, Cloud Storage).
- Designed backend data processing pipelines and Cloud Functions (Node.js) for image transformation, metadata ingestion, and access control, enabling scalable handling of high-volume media and structured data.
- Implemented evaluation-driven workflows for validating annotation consistency and system outputs, improving reliability of human-in-the-loop data labeling systems.
Machine Learning Engineering Intern
Seri Iskandar, Perak, Malaysia
- Developed a scalable code analysis pipeline to detect AI-generated C programs using 9+ software engineering metrics, including Cyclomatic Complexity, Halstead metrics, and entropy-based features.
- Engineered feature extraction and clustering workflows on structured code datasets, achieving a Silhouette Score of 0.48 and identifying distinguishable structural patterns in generated vs human-written code.
- Built automated parsing and preprocessing pipelines for unstructured source code data, enabling downstream ML modeling and evaluation.
Machine Learning Engineering Intern
Remote
- Engineered an incremental learning pipeline in PyTorch to mitigate catastrophic forgetting in evolving classification systems, retaining 85% accuracy after integrating new data classes.
- Evaluated CNN and Vision Transformer (ViT) architectures under continual learning constraints, analyzing trade-offs in knowledge retention, generalization, and model stability.
- Designed experimental evaluation frameworks to benchmark model performance across sequential datasets, improving robustness of model selection decisions.
Data Science Intern
Bangalore, Karnataka, India
- Implemented YOLOv8 architecture to detect alligator cracks on roads, achieving a precision of 91.26%.
- Boosted model recall by 50% using Canny edge detection and CLAHE (adaptive histogram equalization) preprocessing on a 25,000-image dataset.
AI/ML Research Intern
Remote
- Spearheaded VGG19 algorithm implementation, attaining 70% accuracy in symbol and shape recognition using a dataset of 90,000 images.
- Utilized innovative techniques to develop a grid-based display system, achieving an 80% accuracy rate in identifying coordinates for various shapes and symbols.
Skills
Languages
AI & Machine Learning
ML & LLM Systems
Full-Stack & Data
Cloud & DevOps
Projects
Nomz
Django · React · TypeScript · AWS · PostgreSQL · REST APIs · CI/CD
A full-stack restaurant intelligence platform unifying NYC health datasets and user signals across 18+ data models with search, ranking, and CI/CD deployment.
InsightFlow
Python · FastAPI · React · RAG · FAISS · PostgreSQL · Docker · Gemini
A RAG-powered NL-to-SQL assistant that lets non-technical users query relational databases in plain English via a conversational React interface.
Student Performance Visualization
Python · SHAP · UMAP · Scikit-learn
An end-to-end ML pipeline training 10 regression models on student data, with SHAP explainability and UMAP visualization achieving R² of 0.88.
Virtual Mouse
Python · OpenCV · MediaPipe · PyAutoGUI
A real-time computer vision system enabling touchless cursor control through hand gesture and facial tracking using MediaPipe and OpenCV.
Next Word Prediction
TensorFlow · LSTM · NLP · Deep Learning
An LSTM-based language model trained on the Sherlock Holmes corpus for multi-word text prediction using tokenization and sequence generation.
SafeSitePredictor
Scikit-learn · Random Forest · Machine Learning
A Random Forest classifier detecting phishing websites with 96.75% test accuracy across 11,055 samples using URL and HTML feature analysis.
Rejuwel
Python · Tkinter · MySQL · smtplib
A unified desktop health app covering calorie tracking, symptom-to-deficiency mapping, workout planning, and automated daily email health reports.
Historic Stock Forecasting
Python · Pandas · Statsmodels (VAR) · Matplotlib
A multivariate VAR time series model forecasting 5-step closing prices for Apple, Microsoft, Netflix, and Google with ADF stationarity validation.
Phishing Website Detector
Python · Django · Scikit-learn · Pandas · Heroku · SQLite
A Django web app serving a scikit-learn URL classifier in real-time, deployed to Heroku with sub-500ms inference latency and 99% accuracy.
Contact
Let's work together.
I'm actively looking for Summer or Fall 2026 opportunities in software engineering and ML. My inbox is always open.