Angelo
Gaspar
I build machine learning systems that are accurate, explainable, and checked for bias before they ever touch a real decision — trained first as an aeronautical engineer, now calibrated as a data scientist.
Entry-level, PGD-certified, and looking for a Data Scientist, ML Engineer, Data Analyst (ML-focused), or Applied Scientist role.
Angelo Gaspar
Six years ago I was calculating aircraft weight-and-balance to the kilogram, on a deadline, with zero tolerance for error — a Licensed Aeronautical Engineer and Load Control Officer at Macroasia Corporation. That same discipline — verify before you trust, document everything, know exactly how wrong you're allowed to be — is what I now bring to machine learning.
I completed a Post Graduate Diploma in Artificial Intelligence and Machine Learning at the Asian Institute of Management (in partnership with Emeritus) with a 99.6% cumulative average, building full ML pipelines end-to-end: cleaning and imputing real-world data, engineering features, tuning models, and — critically — auditing them for bias before recommending deployment. Alongside the diploma, I've spent since 2019 doing self-directed, data-driven cryptocurrency market analysis, which sharpened the same muscle: turning noisy signals into calibrated, risk-aware decisions.
I'm now looking for an entry-level seat as a Data Scientist, ML Engineer, Data Analyst (ML-focused), or Applied Scientist — somewhere that values rigor as much as results.
Programming & Tools
Machine Learning
Data Foundations
Generative & Applied AI
Analytical & Professional
Real Estate Investment Valuation Engine
House-flipping profitability lives or dies on one question: is the After-Repair Value (ARV) estimate right? Manual Comparative Market Analysis is slow and inconsistent — this engine replaces it with a tuned regression model that prices any listing in seconds, flags underpriced properties, and simulates renovation ROI against the industry's 70% Rule.
- Data: cleaned and imputed a 2,930-record, 80-feature housing dataset — resolved systematically-missing (MNAR) fields with domain-specific rules and removed 5 documented high-leverage outliers, closing 15,732 missing cells to zero.
- Modeling: benchmarked Ridge, Random Forest, and XGBoost via 5-fold GridSearchCV; engineered a Quality × TotalSF interaction feature that SHAP confirmed as the dominant valuation driver. XGBoost won on every technical target (MAPE 6.71% < 8%, R² 0.9517 > 0.90, log-RMSE 0.0914 < 0.12).
- Business logic: a Value Discrepancy Engine and Renovation Simulator flagged 22 undervalued properties from the holdout set, backtesting an 18.9% average ROI per flip.
- Bias audit: ran Demographic Parity, Disparate Impact Ratio, and Equalized Odds checks across neighborhood, zoning, and property-age proxies — found real disparities, then applied reweighting, group-aware thresholding, and post-processing correction. Result: 24 properties flagged, ROI per flip rose to 23.06% (clearing the 20% target), sourcing flag-rate parity improved from a 7.95pp gap to 0.27pp, at a MAPE cost of just +0.06 points.
Seven applied notebooks from the PGD coursework, each a self-contained exercise across the ML lifecycle. Every entry opens the original, fully-run notebook in Google Colab.
Linear regression for housing-price and salary prediction (with residual-plot diagnostics for linearity), plus a logistic regression model predicting employee attrition from HR data.
SVM classifier for customer-segment targeting, and a CNN trained on Fashion-MNIST to classify clothing images across 10 categories — with a look at generalization gaps between training and validation accuracy.
Tuned a classification model with GridSearchCV and RandomizedSearchCV, compared their tradeoffs, and weighed the benefits and risks of leaning on AutoML in production.
K-Means customer segmentation with Elbow-method and Silhouette validation — dropped a low-signal Gender feature to lift mean silhouette score from 0.33 to 0.40, settling on k=6 and giving each cluster a plain-language business interpretation.
Standardized the Wine Quality dataset and applied PCA, plotting explained and cumulative variance to identify the minimum number of components needed to retain 90% of total variance.
Generated a synthetic 365-day time series with trend, weekly seasonality, and noise; decomposed it with additive and multiplicative models, then forecast 60 days ahead using three separate techniques.
Fine-tuned a facebook/bart-base encoder-decoder model on CNN/DailyMail to generate abstractive summaries, evaluated output quality with ROUGE scores, and analyzed how encoder-decoder attention drives content selection.
- Conduct technical and fundamental analysis of financial assets — the same exploratory rigor now applied to EDA on structured datasets.
- Maintain structured transaction records for regular performance evaluation and continuous, data-informed strategy refinement.
- Apply risk-management frameworks to translate quantitative research into decisions under uncertainty.
- Executed precise calculations and data management for aircraft load planning to maximize safety compliance and operational efficiency.
- Delivered accurate, deadline-driven documentation under strict regulatory standards.
- Coordinated cross-functionally with ground, flight, and operations teams — communication skills that transfer directly to remote, distributed data teams.
Post Graduate Diploma in Artificial Intelligence and Machine Learning
Cumulative average: 99.6%. Coursework spanned regression & classification, SVMs & neural networks, model evaluation & hyperparameter tuning, clustering, dimensionality reduction, and generative AI.
B.S. Aeronautical Engineering
CGPA 1.65 / 92.2% — Full Academic Scholar, AY 2012–2013 & 2013–2014. Licensed Aeronautical Engineer, Board Passer at 86.1% (November 2017).
Python for Data Science, AI & Development
Python fundamentals, NumPy/Pandas data analysis, data visualization, foundational ML/AI concepts.
Google AI Essentials
Intro to AI, maximizing productivity with AI tools, prompt-engineering fundamentals, responsible AI use.
Google Prompting Essentials
Advanced prompting techniques, AI-assisted data analysis and presentation building.
Let's talk about the role.
Open to entry-level Data Scientist, ML Engineer, Data Analyst (ML-focused), and Applied Scientist positions — remote or on-site in the Philippines.