J
Juan Kaplan
About
Detail
Buenos Aires, Argentina
Résumé
Jobs
verified_user
0% verified
Al and Data Science InternAssurantSep 2025 - Current (8 months)• Contribute to the full development cycle of ML models, from data pipeline work through to validation and deployment readiness. • Collaborate with cross-functional stakeholders to align technical solutions with business needs. • Build automated, scalable processes to reduce manual overhead across the team.- OAl TrainerOutlier FreelanceOct 2024 - Sep 2025 (1 year)• Evaluated, corrected, and rewrote LLM outputs in mathematics, physics, and programming to produce high-quality supervised training data. • Identified systematic failure patterns in logical reasoning and technical accuracy, contributing to model alignment improvements. • Covered hundreds of evaluations across multiple STEM disciplines over 12 months.
- rLogistics Associater-biopharmApr 2022 - Aug 2022 (5 months)• Managed delivery operations and inventory control, improving shipment accuracy and operational efficiency.
Education
verified_user
0% verified
- Al EngineeringUniversidad de San AndrésJan 2022 - Current (4 years 4 months)GPA 9.48 · 28/42 credits completed
- BHigh School Diploma BilingualBabar Bilingual SchoolJan 2009 - Jan 2021 (12 years 1 month)GPA 9.27
Projects (professional or personal)
verified_user
0% verified
- NOptimizing Ad Placement for Sociodemographic Segments Using Mobility DataNetMob 2025Jan 2025Built a predictive model using anonymized GPS data from Ile-de-France to generate probabilistic heatmaps estimating where sociodemographic segments are likely to be. Pipeline combined GMM per-user spatial modeling, XGBoost on 41 engineered demographic features, and hit@K validation. Achieved Hit@300 coverage of 1.0 for two of three case study segments. Co-authored with researchers from UdeSA and CONICET.
- SStrategies to Avoid Bias in Al-Generated Image DetectionJan 2025Compared fine-tuned ResNeXt101, attention-based models, and a real-image-focused approach with class imbalance correction. The authenticity-focused approach achieved best generalization to unseen diffusion models.
- RRental Price Prediction with Geospatial FeaturesJan 2024End-to-end ML pipeline for Buenos Aires rental price prediction. Compared neighborhood-based vs. proximity-to-POI feature engineering; the proximity approach outperformed on all error metrics.