Lead AI/ML Engineer at Viz.ai

Architected the enterprise rollout of Viz.ai One and Viz Agent Studio on Microsoft Azure, enabling hospitals to deploy guideline-driven clinical pathways significantly faster. Built multimodal healthcare data pipelines (imaging, EHR, operational, patient engagement) using Python, SQL, Kafka, Spark/PySpark, Databricks, Airflow, and Delta Lake. Designed a semantic retrieval layer with Azure AI Search and clinical embeddings, and implemented a knowledge-grounded RAG framework using hybrid retrieval and LlamaIndex to improve retrieval precision. Established model evaluation/adaptation pipelines with Azure Machine Learning, MLflow, PyTorch, and foundation models; added graph-enhanced reasoning with Neo4j. Built production conversational AI and multi-agent workflows (Azure Bot Services, LangGraph, Semantic Kernel), plus clinical NLP and computer vision services. Implemented enterprise MLOps with Azure AI Foundry/AML, CI/CD, Kubernetes/AKS, Terraform, and monitoring, reducing deployment cycle

Lead AI/ML Engineer at Dropbox

Architected the core retrieval platform for Dropbox Dash using Python, FastAPI, Elasticsearch, and AWS services to support indexing and hybrid semantic search across large enterprise document sets. Built data processing pipelines (Databricks, S3) for chunking, metadata enrichment, embeddings, and vector indexing. Implemented hybrid retrieval and semantic ranking combining embeddings with keyword signals and user/document metadata, improving relevance. Developed evaluation/demonstration tooling for RAG pipeline testing, and led LoRA/QLoRA fine-tuning using enterprise search logs to reduce hallucinations. Productionized RAG services with grounded, cited answers via vector search and knowledge-aware workflows. Built low-latency inference services (FastAPI, EKS, Redis caching) and permission-aware retrieval with audit logging. Implemented observability with Prometheus/Grafana/OpenTelemetry/MLflow and built analytics pipelines using Snowflake/BigQuery for retrieval telemetry and LLM evaluat

Senior Machine Learning Engineer at Dropbox

Led ML development for Dropbox Replay, building a distributed speech-to-text platform (Python, Hugging Face, AWS ECS, S3, Kafka) to automate transcription and reduce manual captioning. Implemented NLP-powered comment intelligence to classify/cluster feedback and surface related annotations, improving review efficiency. Built semantic retrieval using Elasticsearch dense vectors, Redis caching, and PostgreSQL metadata indexing for sub-second queries across massive media content. Developed automated media metadata extraction pipelines (Airflow, scikit-learn) and recommendation models (PyTorch) to prioritize feedback and reduce turnaround time. Established end-to-end MLOps with MLflow, Docker, Kubernetes, and CI/CD, improving reliability and reducing release cycles. Added monitoring/observability (CloudWatch, MLflow) and event-driven workflow automation (Kafka, Lambda, Postgres) for indexing and notifications.

Machine Learning Engineer at Wells Fargo

Developed a persona-driven recommendation engine for treasury dashboards and payment workflows using Python, TensorFlow, GraphQL, and behavioral analytics, improving adoption and reducing navigation time. Built cash-flow and liquidity forecasting services using Prophet, XGBoost, and PySpark/Kafka streaming pipelines, improving forecast accuracy and reducing manual effort. Implemented fraud detection/risk scoring models (CatBoost, gradient boosting) and real-time payment anomaly detection (Isolation Forest, Autoencoders, XGBoost) within Kafka-based event processing. Created fraud risk scoring services integrated into approval workflows. Delivered NLP-driven self-service automation with BERT, semantic search, and Elasticsearch. Implemented explainable ML (LightGBM, SHAP, LIME) and experimentation frameworks including multi-armed bandits to optimize user journeys. Contributed to scalable ML pipelines (Airflow, Kubeflow, Docker/Kubernetes, AWS ECS) with drift monitoring, automated retraini

Data Engineer at Sysco

Engineered end-to-end AWS data pipelines (AWS Glue, PySpark, EMR, S3) to centralize sales, inventory, procurement, and logistics data for supply chain analytics. Built scalable ETL workflows and dimensional models in Amazon Redshift using star/snowflake schemas to improve dashboard performance. Implemented CDC-based incremental loading with Glue/Redshift to reduce full-load costs while supporting evolving requirements. Tuned Spark transformations (partitioning, broadcast joins) to reduce batch processing times. Added data lineage/metadata management via Glue Data Catalog and implemented data quality validation/reconciliation in Python/SQL. Created curated data marts for self-service analytics, refactored legacy integrations to serverless Glue with dynamic EMR scaling, and implemented monitoring/alerting with CloudWatch/SNS/SQS to maintain high uptime.

Backend Engineer at Meta

Developed Python/Java backend services supporting high-volume social platform workflows, improving API response time and reliability. Built RESTful APIs with MySQL and Redis, optimized MySQL queries and indexing, and implemented caching to reduce database contention. Engineered Kafka-based asynchronous processing for user-generated events. Improved production stability via monitoring/alerting tools on Linux. Refactored Java components into modular service-oriented applications, and created Hive/Presto reporting workflows for operational analytics. Automated build/testing/deployment using Jenkins and Git to shorten release cycles.

Full Stack Engineer at Google

Contributed to engineering productivity tools, operational dashboards, and reporting platforms. Designed and built Flask-based backend services and REST APIs for infrastructure monitoring, metrics retrieval, and operational analytics. Automated operational workflows and reporting processes to reduce manual effort. Optimized MySQL/PostgreSQL query performance for dashboard responsiveness and integrated deployment telemetry for improved release visibility. Built reusable frontend/backend components to improve maintainability and development velocity.