ML Engineer & Data Scientist — building end-to-end ML systems across CV, NLP, tabular, and astrophysics.https://smooth-cactus0.github.io/https://smooth-cactus0.github.io/projects/birdclef-2026/https://smooth-cactus0.github.io/projects/birdclef-2026/Wildlife species identification from passive acoustic monitoring recordings in the Pantanal wetlands, Brazil. 234 taxa (birds, insects, amphibians, mammals) classified from Mel-spectrograms via EfficientNet, with multi-round pseudo-labeling on unlabeled soundscapes and OpenVINO inference export.Thu, 12 Mar 2026 00:00:00 GMThttps://smooth-cactus0.github.io/projects/jaguar-reid/https://smooth-cactus0.github.io/projects/jaguar-reid/4th place / 348 teams (private LB 0.939, public 0.964). Wildlife re-identification on 1,895 images across 31 jaguar identities. Final pipeline: EVA-02 Large + DINOv2-Large ensemble trained with ArcFace metric learning, followed by pseudo-label fine-tuning (conf > 0.90), query expansion (k=3), and alpha-QE re-ranking. Backbone search spanned EfficientNetV2, ConvNeXt, and ViT before EVA-02 emerged as best single model (+0.069 LB from 10 → 20 epochs).Wed, 11 Mar 2026 00:00:00 GMThttps://smooth-cactus0.github.io/projects/mallorn-astrophysics/https://smooth-cactus0.github.io/projects/mallorn-astrophysics/4th place / 894 teams, Top 0.5% (private LB F1 = 0.6684). Binary classification of Tidal Disruption Events (TDE) vs. non-TDE nuclear transients from simulated LSST lightcurves with extreme class imbalance (~4.8% TDE). Pipeline: Bazin exponential fit parameters + 80+ hand-engineered photometric features → XGBoost + adversarial validation for train/test distribution alignment.Wed, 11 Mar 2026 00:00:00 GMThttps://smooth-cactus0.github.io/projects/gridops-weather-fusion/https://smooth-cactus0.github.io/projects/gridops-weather-fusion/End-to-end ETL pipeline fusing US grid operations telemetry (EIA-930) with NASA POWER hourly weather and solar data across all balancing authorities (2018–2025). Produces an ML-ready Kaggle dataset targeting next-hour load forecasting, peak-load prediction, and anomaly detection.Mon, 09 Mar 2026 00:00:00 GMThttps://smooth-cactus0.github.io/projects/cloud-segmentation/https://smooth-cactus0.github.io/projects/cloud-segmentation/Multi-label semantic segmentation of four satellite cloud patterns (Fish, Flower, Gravel, Sugar) using a ConvNeXt + UNet++ / SegFormer-B3 ensemble with shared classification heads. PyTorch Lightning with 5-fold CV — targeting Dice scores beyond the original 2019 Kaggle competition winners.Sun, 08 Mar 2026 00:00:00 GMThttps://smooth-cactus0.github.io/projects/lidar-roraima/https://smooth-cactus0.github.io/projects/lidar-roraima/Portfolio-grade geospatial ML project on airborne LiDAR tiles over the Roraima-Parima region of Brazil. Two parallel tracks: canopy height modeling (CHM regression) and land-cover classification from LAS point-cloud attributes aggregated into 10 m grid cells.Thu, 05 Mar 2026 00:00:00 GMThttps://smooth-cactus0.github.io/projects/heart-disease/https://smooth-cactus0.github.io/projects/heart-disease/End-to-end clinical ML series: EDA, feature engineering, XGBoost/LightGBM/CatBoost modeling, SHAP explainability, Optuna tuning, and probability calibration on 630K patient records.Sun, 01 Mar 2026 00:00:00 GMThttps://smooth-cactus0.github.io/projects/euclid-morphology/https://smooth-cactus0.github.io/projects/euclid-morphology/ViT-based regression on ESA Euclid Q1 telescope data to predict galaxy morphology parameters. Benchmarks Vision Transformers against Zoobot (Walmsley et al. 2025) on real space-survey imagery.Sun, 15 Feb 2026 00:00:00 GMThttps://smooth-cactus0.github.io/projects/march-mania/https://smooth-cactus0.github.io/projects/march-mania/Bracket prediction for the 2026 NCAA tournament. Engineered 18 features from historical game data, trained XGBoost/LightGBM/CatBoost base models, and used a LogReg meta-learner for calibration. LB score: 0.09451.Sun, 01 Feb 2026 00:00:00 GMThttps://smooth-cactus0.github.io/projects/student-score/https://smooth-cactus0.github.io/projects/student-score/Kaggle Playground S6E1: regression pipeline predicting student exam scores. XGBoost, LightGBM, and ANN base models with Optuna hyperparameter tuning and stacked ensemble.Thu, 15 Jan 2026 00:00:00 GMThttps://smooth-cactus0.github.io/projects/image-forgery/https://smooth-cactus0.github.io/projects/image-forgery/Binary classification of manipulated vs authentic scientific images using EfficientNet and forensic feature extraction. Built for the Recod.ai / LUC Kaggle competition.Mon, 01 Dec 2025 00:00:00 GMThttps://smooth-cactus0.github.io/projects/ariel-exoplanet/https://smooth-cactus0.github.io/projects/ariel-exoplanet/Regression on telescope light curves to predict atmospheric composition of exoplanets. Built for the ESA ARIEL NeurIPS 2024 challenge. Pipeline covers signal preprocessing, feature extraction from multi-wavelength ADC data, and gradient boosting regression.Sat, 01 Nov 2025 00:00:00 GMThttps://smooth-cactus0.github.io/projects/data-analysis-portfolio/https://smooth-cactus0.github.io/projects/data-analysis-portfolio/Five end-to-end projects covering EDA, classification, regression, ETL pipeline engineering, and time series forecasting. Includes customer churn prediction, house price modelling, energy demand forecasting with ARIMA/Prophet/LSTM.Sat, 01 Nov 2025 00:00:00 GMThttps://smooth-cactus0.github.io/projects/nlp-sentiment/https://smooth-cactus0.github.io/projects/nlp-sentiment/Progressive sentiment classification pipeline: Classical ML (TF-IDF + Logistic Regression) → Deep Learning (LSTM) → Transformers (DistilBERT fine-tuning). Demonstrates the performance vs complexity tradeoff across three generations of NLP.Wed, 01 Oct 2025 00:00:00 GMThttps://smooth-cactus0.github.io/projects/retail-inventory-dashboard/https://smooth-cactus0.github.io/projects/retail-inventory-dashboard/Multi-store inventory analytics with ABC classification, demand forecasting, stockout risk scoring, and supplier scorecards. Built with Streamlit for retail operations teams.Wed, 01 Oct 2025 00:00:00 GMThttps://smooth-cactus0.github.io/projects/clinical-trial-dashboard/https://smooth-cactus0.github.io/projects/clinical-trial-dashboard/Phase III clinical trial analytics dashboard covering efficacy endpoints, safety signals, adverse event rates, and survival analysis. Built with Streamlit for medical research teams.Mon, 01 Sep 2025 00:00:00 GMThttps://smooth-cactus0.github.io/projects/ecommerce-dashboard/https://smooth-cactus0.github.io/projects/ecommerce-dashboard/Interactive sales performance dashboard built with Streamlit and Plotly. Features KPI cards, cohort analysis, revenue trends, and one-click PDF reporting for business stakeholders.Fri, 01 Aug 2025 00:00:00 GMThttps://smooth-cactus0.github.io/projects/google-sheets-dashboard/https://smooth-cactus0.github.io/projects/google-sheets-dashboard/Interactive sales performance dashboard built entirely in Google Sheets. Features dynamic KPI cards, pivot-driven charts, dropdown filters, and conditional formatting — no code required to update.Tue, 01 Jul 2025 00:00:00 GMThttps://smooth-cactus0.github.io/projects/isic-skin-cancer/https://smooth-cactus0.github.io/projects/isic-skin-cancer/Multi-modal deep learning pipeline for malignant lesion detection using 3D Total Body Photography. Combines image features with patient metadata for the ISIC 2024 Kaggle challenge.Thu, 01 Aug 2024 00:00:00 GMT