At CIeNET Technologies, I contribute to Google’s open-source MaxText LLM training framework, with 20 merged pull requests. I also built Slurm-based HPC infrastructure on GCP that improved deployment speed by 30% and throughput by 25%.
At Linker Networks, I built Apache Airflow pipelines processing more than 2 million data points per day and a multi-video processing pipeline that cut operational costs by 25%. I also developed backend services and REST APIs for an MLOps platform.
At Influx Finance, I integrated deep learning models into SelfieAuth, improving verification accuracy to 80% and reducing verification errors by 30%. I also researched and implemented image-forgery detection methods.
In my TARA capstone, I built multilingual WMDP benchmark datasets and a resumable evaluation pipeline for LLM research. I handled rate-limited APIs with retries, checkpoints, and per-item cost tracking; the full re-translation cost $4 of a $20 budget.
