S
Sarat Alman
About
Detail
Dallas, Texas, United States
Résumé
Jobs
verified_user
0% verified
Senior AI/ML EngineerClarify HealthMar 2021 - Current (5 years 4 months)• Built and deployed LLM-driven RAG pipelines with LangChain, FAISS, and Pinecone, cutting physician documentation time by 50% across 10M+ EHR records. • Created real-time anomaly detection pipelines with Kafka and Flink, incorporating SHAP explainability for HIPAA/PCI-DSS compliance and ensuring rapid inference latency. • Implemented model observability and drift detection with Prometheus, Grafana, and Datadog, automating retraining triggers via CI/CD pipelines to ensure high uptime. • Deployed predictive models for healthcare resource allocation and patient outcome forecasting using XGBoost, Spark MLlib, and MLflow. • Worked with clinical teams to uphold responsible AI practices, ensuring compliance with HL7/FHIR and KYC/AML standards. No- AI/ML EngineerLumenaltaAug 2018 - Feb 2021 (2 years 7 months)• Fine-tuned LLaMA-3 with LoRA and Hugging Face PEFT on healthcare datasets, deploying on AWS, Azure, and GovCloud via Docker and Kubernetes with REST/FastAPI/GraphQL endpoints. • Developed learning-to-rank systems and hybrid search pipelines (Elasticsearch, vector search, rank fusion) for intelligent document retrieval in insurance and clinical processing. • Built scalable data pipelines and centralized feature stores with PySpark, Airflow, dbt, Snowflake, and Redshift, speeding up model training and real-time inference for multiple teams. • Created and operationalized automated evaluation frameworks and A/B testing infrastructure; tracked experiments with MLflow for model versioning, drift detection, and retraining workflows. • Implemente
- KJunior Machine Learning DeveloperKPG99 INCDec 2016 - Jul 2018 (1 year 8 months)• Built ETL pipelines using Apache Airflow and dbt integrating healthcare data sources into Snowflake and Redshift data warehouses for ML training and downstream analytics. • Developed NLP pipelines (Hugging Face Transformers, spaCy) for sentiment analysis and ticket classification, automating enterprise workflows and improving classification accuracy by 28%. • Supported cloud deployments on AWS Lambda and EC2 with GitHub Actions CI/CD; tracked ML experiments and model versions with MLflow for reproducibility and auditability. • Maintained HIPAA-compliant data pipelines and secure API integrations; wrote complex SQL queries for patient cohort extraction supporting 3 concurrent analytics studies. • Applied supervised and unsupervised ML tech
Education
verified_user
0% verified
- BBachelor of Science in Computer Science
Projects (professional or personal)
verified_user
0% verified
- HHippoRAG: Clinical Trial MatcherMatched unstructured EHRs to clinical trial criteria across 200k patient records using LangChain, GPT-4, Pinecone, and FHIR; lowered patient recruitment time by 60%.
- HHybrid Search & Learning-to-Rank PipelineAssisted hybrid retrieval system (lexical + dense vector, Elasticsearch, reciprocal rank fusion) with LightGBM-based LTR model; instrumented CTR/conversion feedback loops improving nDCG by 22%.
- QQuantized LLaMA-3 (7B) Inference OptimizationApplied GPTQ quantization and deployed on Triton/vLLM; reduced per-token latency from 2s to 300ms on a single T4 GPU, cutting annual GPU costs by $150k.
- EEnterprise RAG Pipeline for Insurance Document ProcessingPlanned scalable RAG pipeline using LangChain, FAISS, and AWS Textract to automate unstructured clinical and insurance document extraction at enterprise scale.
- RReal-Time Fraud Detection System with MLOps ObservabilityDeployed XGBoost ensemble on Kafka/Flink streaming architecture processing 15,000 transactions/sec with SHAP explainability and Prometheus/Grafana dashboards; prevented $2M+ monthly fraudulent volume.
- CClinical AI Agent for Patient Readmission PreventionBuilt a LangGraph-based multi-agent RAG system with EHR integration and human-in-the-loop validation for hospital readmission prevention; lessened readmission prediction latency by 60%.