Senior AI/ML Engineer (Fraud & Forecasting)
Resume:
Carlos
Fonseca
SENIOR AI/ML ENGINEER
CONTACT
Orlando, FL
****************@*******.***
CORE SKILLS
Languages & Data: Python,
Advanced SQL, Pandas, APIs,
Git
ML & Modeling: Scikit-learn,
PyTorch, TensorFlow,
Forecasting, Classification,
Clustering, Anomaly Detection,
Feature Engineering
Stats & Experiments: A/B
Testing, Hypothesis Testing,
Causal Inference, Metric
Design, Cohort Analysis
GenAI & Retrieval: Prompt
Engineering, Hugging Face,
LangChain, LlamaIndex,
Embeddings, RAG Pipelines
Vector Search: pgvector,
Pinecone, Weaviate, Similarity
Search
MLOps: MLflow, Monitoring,
Drift Detection, Observability,
CI/CD, Data Quality Checks
Cloud & Deploy: AWS, Azure,
Docker, Kubernetes,
Containerized Inference,
Infrastructure as Code
Data Engineering: ETL/ELT,
Batch Pipelines, Distributed
Processing, Data Modeling
IMPACT SNAPSHOT
• 9+ years in AI/ML
• 10% fraud detection lift
• 15% forecast error reduction
• 22% fewer reconciliation
errors
EDUCATION
M.S., Computer Science
University of Florida 2015-
2017
B.S., Computer Science
University of Florida 2011-
2015
SUMMARY
Senior AI/ML Engineer with 9+ years of experience in data science and machine learning across banking, retail, healthcare, and enterprise software. Led design and deployment of production ML systems that improved fraud detection by 10%, reduced forecast error by 15%, and supported data pipeline optimization, model performance, and cloud-based AI delivery across Microsoft, U.S. Bank, QuickTrip, and UnitedHealth Group. 9+
Years AI/ML
2x
Deploy speed
30%
Fewer breaks
40%
Faster response
PROFESSIONAL EXPERIENCE
QuickTrip Senior AI/ML Engineer Jun 2023 -
Present
• Devised and implemented an AI-driven forecasting model using 20+ features, including weather proxies and historical sales data, helping cut excess inventory across stores by 8%.
• Instrumented anomaly detection over sales, waste, and replenishment signals, reducing manual exception triage by 6-8 hours per week for operations analysts.
• Hardened operational datasets with automated completeness, freshness, and range checks, lowering downstream dashboard breakages by 30% through proactive alerts.
• Prototyped an internal knowledge assistant using embeddings and vector search, shortening average SOP and incident lookup time from minutes to under 30 seconds. U.S. Bank Senior Data Scientist Nov 2020 –
May 2023
• Engineered transaction and account behavior features using advanced SQL windows and time-based aggregations, improving fraud-risk recall by 8-11% at a fixed false-positive rate.
• Accelerated model deployment cycles by 2x through repeatable evaluation, experiment tracking, and standardized backtesting workflows.
• Pioneered standardized A/B test metrics and decision-ready reports, influencing 9 release choices and identifying the top 5 key user segments.
• Transformed incident response with standardized runbooks and structured logs, accelerating resolution time by 40% and freeing up 10 hours per week. Microsoft Data Scientist Aug 2018 –
Oct 2020
• Analyzed customer behavior patterns to create interpretable segments using clustering and dimensionality reduction, increasing targeted campaign engagement by 6-9%.
• Developed escalation-risk prediction for support cases using text and metadata signals, reducing high-severity escalations by approximately 10%.
• Optimized feature extraction by refactoring CTE-heavy pipelines, indexing critical joins, and minimizing redundant scans, cutting daily pipeline runtimes by 35-50%.
• Deployed containerized inference services with CI/CD checks and observability, improving request latency by approximately 25% through batching and lightweight feature computation. UnitedHealth Group Data Scientist May 2015 –
Jul 2018
• Modeled member risk using claims, utilization, and comorbidity indicators, improving care- management targeting precision by 10-15% versus rule-based selection.
• Automated recurring cohort and cost/utilization reporting with Python and SQL, accelerating leader-facing updates from days to same-day delivery.
• Quantified program impact with hypothesis testing and causal-style comparisons across matched cohorts, supporting expansion decisions against clinical and cost thresholds.
• Catalyzed a 22% decrease in reconciliation errors by implementing standardized ETL processes and data definitions across 30+ critical queries and reports.
Contact this candidate