Principal AI/ML Engineer with a foundation in Biology & Pharmaceutical Sciences. I architect cloud-native ML pipelines, low-latency prediction services, and enterprise data platforms on AWS.
I bridge the gap between biological insight and production-grade engineering.
With an MS in Pharmaceutical Sciences and deep expertise in cloud-native AI/ML, I bring a unique perspective to building intelligent systems. My work spans from molecular-level feature engineering to enterprise-scale data pipelines processing millions of records daily.
I specialize in deploying highly available, security-first ML applications on AWS, building real-time prediction services with sub-100ms latency, and designing data warehouse architectures that power critical business decisions.
MS Pharmaceutical Sciences
BS Biology
Cloud AI/ML on AWS
Real-time ML Systems
Python, AWS, Docker
Snowflake, Airflow
Deploying AI/ML applications in the cloud using Python. Implementing highly available, high-security solutions on AWS. Building data pipelines and daily file processing jobs. Infrastructure as code with Docker & CloudFormation. Application monitoring with DataDog.
Extracting and building AI model features from structured databases and semi-structured data sources using Snowflake and Python. Developing deployment pipelines using Airflow DAG and Step Functions. Identifying model and feature drift using SnowSQL, Python, and SageMaker.
Developing scalable, sub-100ms full stack applications and data solutions using CI/CD methodologies. Resolving production issues across the AI/ML project lifecycle and ensuring end-to-end delivery with monitoring and automated remediation.
Analyzing, architecting, and developing database applications, data warehouses, and operational data stores using ETL tools (SQL, PL/SQL) and Control-M scheduling within UNIX environments. Designing pseudo record keeping systems.
Drug Adverse Event Detection Pipeline
NLP-powered pipeline processing FDA FAERS data to classify adverse drug events by severity. Deployed on ECS Fargate with auto-scaling, S3-based data lake, SQS event routing, and comprehensive DataDog APM monitoring. Full CloudFormation IaC.
Explore ProjectGenomic Variant Classification Service
ML service classifying genetic variants (SNPs, indels) as benign through pathogenic using population frequency, conservation scores, and functional impact features. HIPAA-aware architecture with VPC endpoints, KMS encryption, and WAF protection.
Explore ProjectML Feature Store & Drift Detection
Enterprise feature store built on Snowflake with point-in-time correct retrieval, feature versioning, and lineage tracking. Automated drift detection using PSI, KS tests, and SageMaker Model Monitor with Airflow DAG orchestration.
Explore ProjectDrug-Drug Interaction Prediction
XGBoost-powered pipeline predicting adverse drug interactions using molecular fingerprints, Snowflake-derived patient features, and co-prescription patterns. MLflow experiment tracking with Step Functions orchestration and automated drift-triggered retraining.
Explore ProjectReal-time Drug Response Prediction
Sub-100ms pharmacogenomic prediction API. Predicts patient drug response from genetic profiles using optimized sklearn pipelines, Redis caching, and feature hashing. Includes latency benchmarking, circuit breakers, and full CI/CD with performance gates.
Explore ProjectReal-time Biomarker Monitoring Dashboard
WebSocket-driven clinical dashboard streaming patient biomarker data with ML-powered anomaly detection (Isolation Forest), trend analysis, and clinical alert escalation. Full-stack with real-time charting, CI/CD pipeline, and production monitoring.
Explore ProjectPharmaceutical Data Warehouse
Data Vault 2.0 warehouse for clinical trial data, drug manufacturing, and regulatory submissions. Comprehensive PL/SQL ETL with hub/link/satellite architecture, star schema data marts, Control-M scheduling, and UNIX automation scripts.
Explore ProjectRegulatory Record Keeping System
Pseudo record keeping system for pharmaceutical regulatory compliance. Full audit trail with PL/SQL triggers, submission workflow state machine, cryptographic pseudo-ID generation, and Control-M automated compliance monitoring in UNIX environment.
Explore ProjectMicroscopy Cell Type Classifier
PyTorch CNN classifying blood cell types from microscopy images. Custom CellNet architecture with transfer learning, GradCAM interpretability, stain normalization (Macenko method), and a Streamlit interface for real-time classification.
Explore ProjectAI Drug Molecule Generator
Variational Autoencoder generating novel drug-like molecules from SMILES representations. Property-conditioned generation with Lipinski filtering, QED scoring, PAINS alerts, and latent space interpolation for molecular optimization.
Explore ProjectPlant Disease Detection
Mobile-friendly plant disease detection using EfficientNet-B0 transfer learning. Identifies 15+ diseases across common crops with treatment recommendations. Camera-enabled web interface for field use and a comprehensive disease knowledge base.
Explore ProjectProtein Structure Analysis Tool
Interactive protein sequence analysis with hydrophobicity profiling, secondary structure prediction (Chou-Fasman), Needleman-Wunsch alignment with BLOSUM62 scoring, and disorder prediction. SVG-based visualization dashboard.
Explore ProjectLLM Pharmaceutical Knowledge Assistant
RAG-powered knowledge assistant for pharmaceutical queries using Claude API and ChromaDB. Ingests FDA drug labels and clinical guidelines with semantic chunking, hybrid search, MMR re-ranking, citation attribution, and medical safety guardrails.
Explore ProjectReal-time Pharma Event Streaming
High-throughput streaming pipeline processing prescription and adverse event data in real-time using Kafka and AWS Kinesis. Faust stream processing with sliding-window safety signal detection, S3 Parquet sink, and MSK CloudFormation deployment.
Explore ProjectA/B Testing Platform for ML Models
Full experiment lifecycle platform with Bayesian and frequentist analysis, consistent-hash traffic routing, sequential testing, CUPED variance reduction, SRM detection, and automated champion/challenger model promotion via SageMaker.
Explore ProjectInformatica-Style ETL Framework
Python ETL framework mirroring Informatica PowerCenter patterns: Mappings, Sessions, Workflows, and 11 transformation types (Expression, Lookup, Router, Aggregator, etc.). Includes SCD Type 2 PL/SQL procedures, Control-M jobs, and UNIX automation.
Explore ProjectWildlife Species Classifier
Camera trap image classifier using MobileNetV3 for efficient edge deployment. Identifies 20+ species with IR/night vision handling, biodiversity analytics (Shannon index, occupancy modeling), and AWS Lambda serverless classification.
Explore ProjectAI Recipe & Nutrition Optimizer
Multi-objective recipe optimizer using scipy.optimize to maximize nutrition while preserving taste. USDA nutritional database, ingredient substitution engine, dietary constraint satisfaction, and meal plan optimization.
Explore ProjectTime Series Forecasting for Pharma
Ensemble forecasting system combining ARIMA, Prophet, and ML models for drug demand prediction, shortage early warning, and adverse event trend analysis. Airflow DAG orchestration with automated accuracy monitoring and retraining triggers.
Explore ProjectMulti-Agent AI Research System
Multi-agent system using Claude API where specialized agents (Literature Review, Drug Safety, Medicinal Chemistry, Regulatory Intelligence) collaborate on drug research queries with task decomposition, tool use, and results synthesis.
Explore ProjectTrial Analytics & Biomarker Discovery
Notebook-driven exploratory analysis of a synthetic Phase III anti-inflammatory trial. Baseline balance tables, chi-squared and Mann-Whitney testing, subgroup forest plots, and SHAP-based biomarker discovery with survival analysis via lifelines.
Explore ProjectBeyond core competencies, these emerging skills complement the AI/ML engineering landscape and are areas of active learning and exploration.
Container orchestration for complex ML workloads requiring auto-scaling, GPU scheduling, and multi-model serving across clusters.
Multi-cloud infrastructure as code complementing CloudFormation, enabling hybrid cloud deployments and provider-agnostic resource management.
Domain-specific LLM adaptation through fine-tuning on pharmaceutical corpora, RLHF for safety-critical medical applications, and LoRA/QLoRA for efficient training.
Distributed computing for large-scale feature engineering and model training on petabyte-scale pharmaceutical datasets.
Defending ML models against adversarial attacks, data poisoning, model extraction, and ensuring compliance with AI governance frameworks.
Building internal ML platforms with self-service model deployment, feature stores, experiment tracking, and automated retraining orchestration at scale.