Luis Miguel Montoya Henao
About
Detail
Antioquia, Colombia
Résumé
Jobs
verified_user
0% verified
Senior Data Scientist public Remote experienceUPLabsApr 2026 - Current (7 months)• Build the AI core of PartsPulse, an AI-native operations platform for aftermarket parts businesses (first startup from UP.Labs' Manufacturing Venture Lab with Wabash), unifying inventory planning, pricing optimization, and sales intelligence in one system.
• Design data engineering and ML pipelines on Databricks that ingest sales, customer, and inventory data and turn them into automated operational decisions: anomaly flags, pricing recommendations, stocking signals, and prioritized sales actions.
• Work directly with founders to convert business problems into production ML and LLM-assisted workflows, replacing manual analysis across pricing, inventory, and dealer operations.- Data Scientist EngineerMercado LibreNov 2024 - Apr 2026 (1 year 6 months)• Built an LLM-powered review workflow for risky transactions end to end: defined the intent taxonomy, curated seed data, and shipped an offline evaluation harness with human review in the loop, automating triage that previously required full manual casework. Orchestrated end-to-end REST API calls for LLM providers, internal data services, and review tooling, initially using n8n webhook-triggered flows before migrating to a custom harness.
• Shipped a real-time anomaly detection system on credit time-series features that surfaced ~70% of emerging account-takeover and device-theft patterns without human monitoring. Served and consumed prediction endpoints over REST APIs and containerized training/inference environments with Docker to en - Data ScientistSistecréditoMay 2024 - Nov 2024 (7 months)• Automated the credit-risk workflow on Azure Databricks: ETL pipelines, scheduled retraining, and MLflow tracking, cutting analytics turnaround from manual runs to a hands-off cycle.
• Predicted defaults at 80% accuracy and built payment-behavior segmentation that fed automated collection and recovery decisions.
• Used LLMs as a first-pass layer in the data pipelines to categorize and filter messy text fields, improving data quality. - Development SpecialistSeguros SURANov 2023 - May 2024 (7 months)• Priced corporate health plans with predictive risk models, balancing premium hikes with retention.
• Delivered interactive dashboards and analytics tooling adopted by Sales and Portfolio teams.
• Leveraged database systems to optimize the pricing system for corporate health insurance plans, utilizing health and finance datasets to enhance decision-making processes.
• Employed distributed computing techniques to manage large datasets efficiently, ensuring timely analysis and reporting.
• Developed an A/B testing framework to validate pricing strategies, ensuring that changes did not negatively impact user satisfaction or key business metrics.
• Utilized cloud services for scalable data storage and process - Data ScientistSistecréditoJul 2023 - Nov 2023 (5 months)• Developed LSTM default-prediction and customer-lifetime-value models, mitigating risk and boosting marketing ROI. • Automated data pipelines in Azure Databricks, enabling faster analytics turnaround.
- Used LLMs as a first-pass layer in the data pipelines to categorize and filter messy text fields, improving data quality. - LFinancial Data Analyst & InternLínea DirectaAug 2022 - Jul 2023 (1 year)• Applied Double Machine Learning to set dynamic prices and launched the team's first MLOps framework, which introduced Git-based versioning for machine learning models. Built and deployed REST API prediction services consumed by business systems.
• Improved sales-forecast accuracy 10% using a Temporal Fusion Transformer, enhancing financial planning. Implemented containerized environments using Docker to ensure dependable scheduled retraining and pinned dependencies.
Education
verified_user
0% verified
- M.S. in Computer ScienceGeorgia Institute of TechnologyAug 2025 - Current (1 year 3 months)
- TTOEFL iBT 110Jan 2025
- TTensorFlow Developer Professional CertificateFeb 2024
- DDeepLearning.AI Deep Learning SpecializationFeb 2024
- GGoogle Data Analytics Professional CertificateMay 2023
- UB.S. in PhysicsUniversidad EIAJul 2019 - Jul 2023 (4 years 1 month)
Projects (professional or personal)
verified_user
0% verified
- OOpen Source & Research InfrastructureSep 2025 - Current (1 year 2 months)Maintain public open-source repositories on GitHub, including the released code for a first-author ICML 2026 workshop paper. Implement CI/CD, tagged releases, and reproducibility standards. Develop and maintain open-source tools including Rundial, Podmon, and the interface layer of GamesBench, an LLM benchmark. Utilize Docker to package experiment environments for reproducibility on GPU nodes and build resilient API clients for OpenAI, Anthropic, and OpenRouter.
- UEstimating Price Elasticity of Demand in the Fashion Industry through Causal Inference.Universidad EIAAug 2022 - May 2023 (10 months)Estimated price elasticity of demand for fashion retail products using causal inference rather than naive correlational methods. Applied Double Machine Learning to isolate the causal effect of price on demand from confounders (seasonality, promotions, product mix), producing elasticity estimates reliable enough to drive dynamic pricing decisions. The approach addressed the core problem of observational pricing data: prices are set in response to demand conditions, so standard regression conflates the price effect with everything that moved prices in the first place.
- ULanguage Models Applied to Venture Capital Funding Prediction: Evidence for Latin America.Universidad EIAJul 2022 - Jun 2023 (1 year)Bachelor's thesis (Physics, Universidad EIA). Applied language models to predict venture capital funding outcomes for Latin American startups, using the text that describes a company [pitch/profile descriptions, from e.g. Crunchbase or a comparable source] as the predictive signal alongside structured features. The regional focus was the point: Latin American startups are underrepresented in funding data and in the tools built on it, so the thesis tested whether language-model representations could extract signal in a data-scarce market where standard quantitative screening has little to work with. Covered [data collection and cleaning, model comparison against classical baselines, and evaluation of predictive performance].
- UPredicting Deforestation via Deep Learning and Satellite ImagesUniversidad EIAJan 2021 - Jul 2021 (7 months)Built an end-to-end deep learning pipeline to predict future deforestation from satellite imagery. Constructed a spatial grid over the study region, assembled a multi-year timeline of observations per grid cell, and trained a U-Net to segment which cells would be deforested in the coming period. Covered the full workflow: satellite data acquisition and preprocessing, spatiotemporal dataset construction, model training and evaluation, and map-level visualization of predicted risk zones.
- FFRC Robotics CompetitionFRCJan 2017 - May 2020 (3 years 5 months)Co-led the software team of a FIRST Robotics Competition (FRC) team for three years, taking a robot from concept to competition each season under FRC's six-week build deadline. Owned the programming side of the robot: autonomous routines, driver-control systems, and sensor integration, coordinating the software workstream with the mechanical and electrical teams and onboarding new members each season.