Conducting code reviews for student implementations of scalable ML pipelines, focusing on memory management and concurrency.

Collaborated with faculty to design assignments on real-world ML workflows, including data preprocessing pipelines and model evaluation metrics.

Wrote Python scripts to automate grading of Jupyter notebooks, reducing feedback turnaround time by 40%.

Architected a Research Discovery Engine using LangGraph and Llama 3.1 70B to map global researcher collaborations by indexing massive datasets from Elsevier (ScienceDirect/Scopus).

Implemented a Multi-Agent ”Search Router” to handle hybrid semantic queries across scientific nomenclature; utilized BGE-M3 and Reciprocal Rank Fusion (RRF) to improve keyword matching accuracy by 40%.

Engineered a hybrid cloud processing pipeline using Node.js and FastAPI on AWS ECS, scaling asynchronous workers to sustain 99.9% uptime under peak loads of 3K+ requests per second.

Optimized Retrieval Performance by deploying a Write-Through Redis cache, slashing P99 latency for complex ”Chain of Retrieval” lookups from 450ms to <100ms.

Led Full-Stack Ownership of the application lifecycle, designing the CI/CD strategy over Kubernetes and persistent volumes to reduce deployment lead times by 65%.

Built a FastAPI microservice on AWS ECS (Fargate) utilizing AsyncIO to parallelize data fetching from 5+ external APIs, slashing P99 latency from 450ms to <100ms.

Deployed a Redis Write-Through caching layer with custom TTL eviction policies, handling >300 RPS with sub-millisecond retrieval times.

Modernized CI/CD pipelines using Docker and Kubernetes, enabling zero-downtime blue-green deployments.