I build enterprise-scale distributed systems, real-time inference pipelines, and production software. My work spans 24B+ records/day at sub-200ms latency, with $122.9M+ in quantifiable cost savings through ML-driven optimization.
%20Intern%20%7C%20Summer%202026-FF9900?style=for-the-badge&logo=amazonwebservices&logoColor=white&labelColor=232F3E){kind=icon}
Graduate Research Assistant, University of Maryland - Sep 2026-present HPC I/O research for Open OnDemand under Dr. Alan Sussman, in collaboration with the Ohio Supercomputer Center and Princeton Plasma Physics Lab (PPPL).
Software Engineer (ML) Intern, AWS Redshift ML Team - Summer 2026 (completed) Shipped a native vector search prototype into Redshift's query engine (500K-50M vectors, sub-100ms) and Bedrock ML integrations that cut LLM time-to-production from weeks to days. Won the org-wide Amazon AUTA AI Hackathon among 800+ Amazon interns for a multi-model cost-aware router (3.2x cheaper inference). Case study →
GSoC 2026 Contributor - Hardened the Accord Project Agreement Protocol MCP server under The Linux Foundation. Shipped the typed service-layer port, MCP SDK 2.0 migration, PG18 + RLS smoke, paged MCP resource URIs, and the A2A wrapper design of record. Mentors: Niall Roche, Dan Selman. Blog write-up →
| Domain | Project | Highlights |
|---|---|---|
| AWS / Redshift ML | AWS Redshift Vector Search + Bedrock ML | Native semantic search in Redshift query engine (500K-50M vectors, sub-100ms); Bedrock ML integrations cut LLM time-to-production weeks→days. Amazon AUTA Hackathon Winner (org-wide, 800+ interns, 3.2x cost reduction). |
| Open Source / GSoC '26 | Accord Project APAP/MCP | Hardened the MCP server: typed service layer, MCP SDK 2.0 migration, PG18 support, A2A wrapper design. POC · Blog |
| GenAI / LLM | Production LLM Serving Framework | vLLM continuous batching, 12.3K req/sec, 42ms P50, INT4 quantization, 70% memory savings |
| Multi-Agent AI | ConstellationAI Multi-Agentic RAG | LangGraph orchestration, semantic caching, real-time token optimization, GPU monitoring |
| Data Engineering | Enterprise Data Warehouse | 24B records, <60s latency, 100K records/sec, 32 fact tables, 128 dimensions |
| Distributed Systems | Big Data Query Platform on K8s | Distributed query engine on Kubernetes, horizontal scaling, fault-tolerant execution |
| Medical Imaging | Real-Time Medical Image Enhancement | DDPM + 3D U-Net for CT/MRI, SSIM: 0.89, <2s on 512³ volumes, 92% radiologist approval |
| Quantitative Finance | Quant Trading Platform | LSTM/Transformer HFT, Sharpe: 2.1, 87.3% directional accuracy, 1M+ ticks/sec |
| Computer Vision | Industrial Quality Control CV | YOLOv8 + ResNet-50, 94.2% accuracy, 500+ images/min, <150ms, TensorRT |
| Infrastructure | Automated Infra Provisioning | Terraform, Ansible, Puppet, VMware vSphere/NSX-T, 85% faster deployments, 95% compliance |
| Experimentation | Real-Time Experimentation Platform | Thompson sampling, CUPED, causal inference, Benjamini-Hochberg, 50+ experiments |
Languages · Python · Java · Scala · SQL · TypeScript · JavaScript · C++ · Go · R
ML & AI - PyTorch · TensorFlow · Scikit-Learn · HuggingFace · LangChain · and more
PyTorch · TensorFlow · Scikit-Learn · XGBoost · LightGBM · HuggingFace Transformers · LangChain · LangGraph · OpenAI · Anthropic Claude · MLflow · Weights & Biases · Optuna · Ray Tune · ONNX · TensorRT · vLLM · NVIDIA NeMo · Stan · PyMC
Backend & APIs - FastAPI · Flask · Spring Boot · Node.js · and more
FastAPI · Flask · Django · Spring Boot · Node.js · Express · REST · gRPC · GraphQL · WebSockets · Celery · RabbitMQ · API Gateway
Data & Distributed Systems - Spark · Kafka · Airflow · Snowflake · and more
Apache Spark · Apache Kafka · Apache Airflow · Prefect · Apache Flink · Delta Lake · Apache Hudi · Databricks · Snowflake · dbt · Apache Hive · Presto/Trino
Databases & Storage - PostgreSQL · MongoDB · Redis · Neo4j · and more
PostgreSQL · MySQL · MongoDB · Redis · Elasticsearch · Apache Cassandra · Neo4j · ClickHouse · DynamoDB · Redshift · BigQuery · Pinecone · ChromaDB · S3 · HDFS
Cloud & Infrastructure - AWS · Azure · GCP · Kubernetes · Terraform · and more
AWS: SageMaker · Bedrock · Lambda · Step Functions · EKS · ECR · EventBridge · Glue · Kinesis · Redshift · S3 · CloudFormation · CDK Azure: Data Factory · Synapse · Purview · AKS · Azure ML GCP: Vertex AI · BigQuery · GKE · Dataflow · Pub/Sub IaC & Containers: Terraform · Ansible · Puppet · Docker · Kubernetes · Helm
DevOps & Monitoring - GitHub Actions · Jenkins · Prometheus · Grafana · and more
GitHub Actions · Jenkins · ArgoCD · GitOps · Prometheus · Grafana · ELK Stack · Datadog · CloudWatch · PagerDuty · SonarQube · Trivy · CI/CD pipeline design
Analytics & Visualization - Tableau · Power BI · Plotly · Streamlit · and more
Tableau · Power BI · Apache Superset · Looker · Plotly · Streamlit · Jupyter · Matplotlib · Seaborn · D3.js
Statistical Methods - Bayesian Inference · Causal Inference · Survival Analysis · and more
Bayesian Inference (Stan/PyMC) · MCMC · Hierarchical Models · Causal Inference · Propensity Score Matching · Difference-in-Differences · Instrumental Variables · Survival Analysis (Cox/Kaplan-Meier) · Time Series (ARIMA/GARCH/State-Space) · A/B Testing · Multi-Armed Bandits · Sequential Testing · Power Analysis
| 24B+ records/day processed |
<200ms real-time inference |
$122.9M+ ML-driven savings |
13.4K crashes prevented |
| 99%+ system uptime |
74.4% infra cost reduction |
2 IEEE + 1 IJAET peer-reviewed publications |
130+ professionals mentored |
- Amazon AUTA AI Hackathon Winner - Summer 2026 - org-wide top winner among 800+ Amazon interns for a multi-model cost-aware router (3.2x cost reduction on production inference)
- Google Summer of Code 2026 - Selected contributor for Accord Project (The Linux Foundation), Idea #4: Hardening the APAP / MCP Server
- Best Performer FY 2023-24 - Bridgestone Group, recognized for enterprise AI delivery
- Hackathon Judge & Mentor - HackMIT, UCLA LA Hacks
- Peer-Reviewed Publications - 2 IEEE papers and 1 IJAET journal on ML for predictive maintenance and supply chain optimization; viXra preprint on automotive telematics
- Predictive Maintenance in Automotive using Machine Learning - IEEE Digital Xplore Library (Accepted, September 2024; ISSN 2169-3536; Manuscript Access-2024-36520). LSTM + SHAP interpretability on high-frequency CAN bus telematics; distributed training on Apache Spark, deployed on AWS Lambda for real-time failure prediction.
- Optimizing Supply Chain Management Through Data Science and AI: A Data-Driven Approach - IJAET, Vol. 5 No. 4 (December 2023; ISSN 2633-4828). Hybrid descriptive + prescriptive analytics framework for supply chain optimization; predictive lead time, inventory turnover, and demand variability forecasting.
- Predictive Maintenance in Automotive Telematics using Machine Learning - viXra:2511.0028 (preprint). K-NN, Random Forest, SVM for truck load classification via CAN bus telematics; 95.97% accuracy.