Architecting resilient data ecosystems and scalable software systems.

MS Computer Science student at Northeastern University (4.0 GPA) graduating in December 2026. I specialize in designing scalable distributed systems, cloud-native lakehouses, and production-grade pipelines. Beyond simply connecting modern tools, I am deeply committed to building and understanding the foundational architecture of the systems I engineer.

🎯 Actively seeking Fall 2026 Internships/Co-ops and Full-Time opportunities in Data Engineering, SWE/SDE, Analytics Engineering, and BI.

LinkedIn Email

• Languages: Python, SQL, Java, Bash
• Big Data & Streaming: Apache Airflow, Apache Kafka, Apache Flink, Apache Spark (PySpark), dbt, Great Expectations, Delta Lake, Parquet
• Databases: PostgreSQL, MySQL, Snowflake, Redis, TimescaleDB, DuckDB, Cassandra, DynamoDB
• Cloud & DevOps: AWS (S3, EC2, Glue, IAM, Redshift, CloudWatch), Azure (Databricks, Data Factory, Data Lake), Terraform, Docker, Kubernetes Git, GitHub Actions, CI/CD

May 2026 - Present

• Held weekly office hours debugging student Python implementations of PCA, multiple linear regression, and Ridge/Lasso, working through algorithm internals and scikit-learn pipelines with a graduate cohort.
• Graded course assignments on a 10 to 12 day turnaround, reviewing model code, train/test splitting logic, and written analyses against the course rubric.

Sept 2025 — Present

• Engineered a Composite Semantic Drift Score for a co-authored COLM 2026 paper on LLM paraphrasing, as measured by cumulative meaning loss exceeding 331% of safety thresholds across 36,827 records, by integrating SBERT, METEOR, and ROUGE-L signals into a single weighted index.
• Architected automated evaluation pipelines in Python to process 4,817 complex records, engineering batched scoring mechanisms that replaced manual analysis workflows.
• Eliminated missing-field errors and achieved 100% data completeness by enforcing strict Pydantic schema contracts and staged quality gates across multi-modal data ingestion pipelines.
• Optimized embedding throughput and reduced pipeline runtime to under 75 seconds by implementing all-mpnet-base-v2 batch encoding and fingerprint-based cache reuse.
• Quantified non-linear semantic drift across multi-hop text generation chains, as measured by a Hop A to Hop B t-statistic of 213.15, by running paired t-test and Wilcoxon signed-rank validation scripts over the full evaluation set.
• Engineered a scalable multi-model ingestion matrix utilizing directory-driven loaders and metadata injection, automating data processing across 7 distinct domains without manual intervention.

• Tech: Java, RabbitMQ, Redis, MySQL, WebSockets, AWS EC2
• Impact: Engineered a write-behind persistence pipeline sustaining throughput of 21,091 msg/s with zero data loss. Architected CQRS-style read/write separation and optimized read-path latency to 13ms at 1M-row scale.

• Tech: Terraform, AWS S3, Glue, Airflow, PySpark, dbt, Docker
• Impact: Processed 100GB+ NYC taxi trip records (2.8M rows) through PySpark ETL on AWS Glue. Provisioned infrastructure using Terraform IaC and automated daily batch pipelines via Airflow DAGs.

• Tech: Kafka, Flink (Java), Redis, PostgreSQL, Docker
• Impact: Achieved 99% polling reduction via Kafka key-based partitioning and Flink exactly-once processing. Architected a hybrid Redis/TimescaleDB dual-storage system serving 20+ concurrent users with sub-second response times.

• Tech: Python, PostgreSQL, Snowflake, Airflow, Docker, marimo
• Impact: Designed a Medallion-architecture warehouse (Bronze → Silver → Gold) centralizing 14 sources for 1.6M+ records. Reduced SQL query latency by 90% via query tuning and data normalization.

• Tech: Apache Airflow, dbt, PostgreSQL, AWS, Terraform, Docker
• Architecture: S3 Data Lake → dbt transformations → Analytics Mart with SCD Type 2 dimensions
• Impact: Optimized query time from 4.2s → 1.1s (74% improvement) across 3 data sources and 50K+ events.

• 📧 [email protected]
• 💼 LinkedIn
• 📍 Seattle, WA

I believe in building things the right way — production-grade code, proper documentation, and solutions that actually work.