About
I was born in Houston, Texas to Argentine immigrants and grew up in Bentonville, Arkansas before heading to Boston, Massachusetts for college at Harvard University where I concurrently completed my bachelor's and master's degrees in Computer Science.
During my time at Harvard, I worked in the Ability Lab on assistive devices. I also worked as a data scientist at Massachusetts General Hospital for neurological research. After my third year, I had the opportunity to intern as a software engineer in San Francisco, California at Netflix.
I'm passionate about software engineering, startups, systems, security, and AI/ML. I love building things, playing with AI tools, and solving problems.
In my free time I enjoy running, playing the viola, and sports, especially MMA and soccer. Feel free to reach out if you want to collaborate on anything or just to chat!
+1 (479) 531-3651 | Boston, Massachusetts, United States
Languages
-
Python -
Java -
JavaScript -
TypeScript -
HTML -
CSS -
Go -
Rust -
C++ -
C -
C#
Technologies
-
React -
Spring Boot -
Next.js -
Node.js -
PyTorch -
TensorFlow -
OpenCV -
Git -
AWS -
Docker -
GraphQL -
gRPC -
Jenkins -
PostgreSQL -
MongoDB -
Firebase -
Supabase
Education
Harvard University
Cambridge, MA Master of Science in Computer Science August 2025–May 2026
Harvard University
Cambridge, MA Bachelor of Arts in Computer Science with Honors August 2022–May 2025
- Coursework: Data Structures & Algorithms, Operating Systems, Distributed Systems, Networks, Databases, Systems, Machine Learning, Deep Learning, Probability & Statistics, Linear Algebra, Computer Vision, Compilers, Optimization, Artificial Intelligence
Experience
Netflix
Los Gatos, CA Software Engineer Intern, Game Systems June 2025–August 2025
- Created GraphQL mutation endpoint using Java and Spring Boot on enterprise edge, integrating with gRPC services to verify test accounts, check push consent, and dispatch test notifications supporting 1,000,000+ daily requests with 99.999% availability
- Added two new fields to existing GraphQL query, fetching data via gRPC, and displaying results with React and TypeScript
- Built full-stack push consent dashboard, adding two GraphQL query and mutation endpoints managing 800,000+ daily requests
- Developed full-stack notification interaction tracking system with five new GraphQL endpoints handling 600,000+ daily requests
Harvard Ability Lab
Allston, MA Software Engineer September 2023–May 2024
- Integrated a mobile computer vision app with embedded hardware to make a self-steering white cane system to avoid obstacles
- Built a self-steering system with a microcontroller, encoder, and motor using SPI communication and PID control in C++
- Implemented Unity script in C# to transmit obstacle and path data from the app to the microcontroller via Bluetooth for steering
Massachusetts General Hospital
Boston, MA Data Scientist Intern, Global Neurology Research Group August 2023–July 2024
- Created data pipeline from patient registry with SQL and Python reducing dimensionality 83% retaining 90% variance with PCA
- Developed polynomial regression model with 94% accuracy in predicting MRI lesion counts enabling 3x faster training
- Co-authored peer-reviewed paper in Neurological Sciences on effects of GLP-1 agonists (e.g., Ozempic) on MS in 49 patients
Publication: Udawatta, M., Fidalgo, N. & Mateen, F.J. Multiple sclerosis patients taking glucagon-like peptide-1 receptor (GLP-1) agonists: a single-institution retrospective cohort study of tolerability and weight loss. Neurol Sci 46, 343–349 (2025).
Projects
- Co-founded and engineered an AI-powered SAT preparation platform serving 10+ paying customers with a 4.9/5 user rating
- Built full-stack web app featuring dynamic skill tracking, personalized quick practice sessions, and real-time AI tutor chatbot
- Created 2,000+ question bank and 15+ full-length practice tests with question and domain typing to enable targeted practice
- Built microservices platform for generating videos from natural language using LangGraph agents and RAG with ChromaDB.
- Integrated full-stack app with React/Vite frontend, FastAPI backend, Firebase authentication, and GCS media storage, deployed to GCP using Kubernetes, Pulumi, and Docker Compose orchestration across 4 containerized services in CI/CD pipeline.
- Implemented retry system with RAG, diff-based video editing, and TTS podcast generation with automated captions.
LSM-Trees with Machine Learning
GitHub Python, PyTorch, NumPy, scikit-learn
- Designed classifier algorithm with gradient boosted trees reducing query latency by 2.3x and 30% fewer Bloom filter checks
- Built Bloom filters with ML and lightweight backup filters reducing memory footprint 70–80% per level and zero false negatives
- Trained and cross-validated models on synthetic and real key-value workloads achieving up to 91% accuracy on level prediction
Publication: Fidalgo, N. Ye, P. (2025). Learned LSM-trees: Two Approaches Using Learned Bloom Filters.
Simulating Evolvability as a Learning Algorithm
GitHub Python, NumPy, Matplotlib
- Conducted first empirical study of evolvability for six Boolean function classes across four distributions with a genetic algorithm
- Discovered majority function is evolvable under uniform, binomial, and biased Bernoulli distributions but not beta distribution
Publication: Fidalgo, N., Ye, P. (2025). Simulating Evolvability as a Learning Algorithm: Empirical Investigations on Distribution Sensitivity, Robustness, and Constraint Tradeoffs. arXiv:2507.18666.
High-Performance LSM-Tree Storage Engine
GitHub C++
- Designed LSM-tree with skip list memtable, variable false positive rate Bloom filters, and hybrid compaction strategy
- Achieved sub-linear latency scaling from 100MB–10GB data and up to 40% higher write throughput under skewed workloads
- Demonstrated near-linear scalability to 16 threads and 32 concurrent clients, reducing latency 12x and increasing throughput 25x
Publication: Fidalgo, N. (2025). Design and Implementation of an LSM-Tree Storage Engine.
Extending U-Net for Semantic Segmentation
GitHub Python, PyTorch, OpenCV, NumPy, Matplotlib
- Evaluated U-Net with residual blocks and batch normalization and hybrid fully convolutional network on CamVid urban dataset
- Improved dominant class accuracy (0.954 Dice score for sky and 0.928 for road) with residuals and combined loss functions
Publication: Fidalgo, N., Nair, J. (2025). Implementing and Extending U-Net for Semantic Segmentation.
Multimodal AI for Forensic Sketch Generation
GitHub Python, PyTorch, CUDA, Hugging Face
- Achieved 21% higher structural similarity and 25% higher peak signal-to-noise ratio over Stable Diffusion v1.5
- Fine-tuned CLIP model on attention heads using LoRA improving text-sketch alignment by 9% and reducing perceptual error 2%
Publication: Fidalgo, N., Contreras, A., Harvey, K., Ni, J. (2025). Gen-AI Police Sketches with Stable Diffusion. arXiv:2507.18667.
Links