Conversational AI
Real Estate Chatbot
Built a WhatsApp assistant for real estate inquiries using Twilio, Flask, BERT, LangChain, and RAG to deliver accurate property information and reduce manual workload for agents.
Data Scientist · New York City
Turning data into clear stories, useful systems, and better decisions.
I’m Juan Acosta, a data scientist with a background in music and a focus on machine learning, NLP, and urban analytics. I like building work that feels rigorous, human-centered, and practical.
About
A cross-disciplinary path into data science.
My journey from music teacher to tech professional shaped the way I approach data: analytically, collaboratively, and with a strong sense of structure. I specialize in data acquisition, modeling, machine learning, and NLP, with a particular interest in urban planning and transportation.
Outside of work, I’m still deeply connected to jazz and rhythm. I also enjoy soccer, coffee, chess, and the broader question of how data and AI can be used responsibly for positive change.
Selected Projects
Work grounded in real use cases, automation, and applied modeling.
Audio ML
Urban Sound Classifier
Trained and compared machine learning models on extracted audio features to classify urban sounds, with XGBoost reaching about 70% accuracy.
Deep Learning
Speech Recognition
Designed a voice recognition model for 10 individuals using Mel-spectrograms, MFCCs, and bidirectional LSTMs, reaching 85% accuracy.
Data Infrastructure
Distance Matrix
Created a pipeline around the Google Maps Distance Matrix API to collect, structure, and store travel data for fast use in Python, Pandas, and SQL workflows.
Writing
Explaining technical ideas in a way that stays accessible.
Building a Real Estate WhatsApp Chatbot with AI and Automation
A walkthrough of building an automated WhatsApp chatbot with Twilio, Flask, BERT, LangChain, and retrieval-augmented generation.
Data-Driven Urban Planning: Polar Histograms and OSMnx
An exploration of urban informatics using OSMnx and polar histograms to better understand transportation networks and city structure.
Using Google Maps Distance Matrix API to Create a Distance Table
A practical guide to calling the Google Maps API, shaping the results with Pandas, and storing them in SQLite for efficient reuse.
Unveiling Karatsuba Multiplication
A mathematical and code-focused look at Karatsuba multiplication and why it matters for algorithmic efficiency.
Capabilities
A toolkit spanning data work, modeling, experimentation, and deployment.
Core
- Python
- SQL
- Machine Learning
- Data Analysis
- A/B Testing
- NLP
Tools & Platforms
- MySQL, SQLite
- Pandas, NumPy, Jupyter
- Scikit-Learn, TensorFlow, Keras
- XGBoost, PyTorch, NLTK, BERT
- AWS: EC2, S3, Lambda
- REST APIs, Postman, Flask
Specialties
- Data Visualization with Matplotlib, Seaborn, Plotly, Tableau
- Web Scraping with BeautifulSoup
- Google Maps API integrations
- Audio processing with Librosa
- Git and GitHub workflows
Education
Technical training supported by a deep background in music and performance.
NYU Tandon School of Engineering
Completed the NYU Tandon Bridge program, a rigorous foundation in computer science and mathematics designed to prepare students for advanced study in engineering and computing.
Flatiron School
Completed intensive training in data acquisition, modeling, statistical analysis, machine learning, deep learning, and natural language processing.