Aayush Jha

Problem

Unstructured CI/CD build logs make it difficult to quickly isolate root causes of pipeline failures.

Solution

Engineered a data pipeline using Python and Regex to automatically extract, parse, and tokenize unstructured build logs into structured datasets. Trained a machine learning classification model using pandas and scikit-learn.

Impact

Automated root-cause error isolation and predicted pipeline failures based on log patterns.

Technologies

• Python
• Pandas
• Scikit-Learn
• Regex
• Machine Learning

Problem

Need for a virtual file system that merges read-only and read-write directory layers with full POSIX operation support, replicating Docker’s layered filesystem model.

Solution

Implemented a simplified Union File System in C using FUSE, merging read-only and read-write directory layers into one virtual file system with full POSIX operation support. Engineered Copy-on-Write semantics and whiteout-based deletion handling; validated against an automated test suite.

Impact

Successfully replicated Docker’s layered filesystem model with robust POSIX support and verified against an automated test suite with all tests passing.

Technologies

• C
• FUSE
• Linux (Ubuntu)
• POSIX APIs

Problem

High computational latency and bottlenecking when processing large 4K high-resolution images sequentially.

Solution

Developed a distributed image processing pipeline using Apache Kafka and Python across 5 worker nodes. Split 4K images into 64 chunks to achieve parallel processing and dynamic load distribution.

Impact

Reduced image processing time by 70% and achieved a 5x throughput improvement through parallelized chunking.

Technologies

• Apache Kafka
• Python
• Distributed Computing