# Professional Experience

## Software Engineer - AI Platform | Cloud Systems LLC | US Remote

**Jul 2025 - Present**

- **Agentic Workflow Automation**: Architected an agentic workflow using **LangGraph** and **ReAct** to automate SQL generation. This system automated **~65%** of ad-hoc data-auditing requests from internal stakeholders, reducing the average response time from **4 hours to under 2 minutes**.
- **ETL Optimization**: Optimized data ingestion performance by rebuilding ETL pipelines with batched I/O, incremental refresh logic, and dependency pruning, cutting daily execution runtime by **25%**.
- **Infrastructure & Reliability**: Improved production reliability by shipping the agent service on **Kubernetes** with autoscaling and rolling deploys, adding alerts and rollback steps for failed releases.
- **Contract Testing**: Improved cross-service reliability by implementing **Pydantic** schema validation and contract tests, preventing multiple breaking changes from reaching production.

## Machine Learning Engineer | Virginia Tech, Dept. of Plant Sciences | Blacksburg, VA

**Aug 2024 - Jul 2025**

- **Model Optimization**: Increased genomics sequence classification throughput by **32%** by applying **LoRA** and **soft prompting** methods. Packaged repeatable **PyTorch** pipelines that cut per-experiment training time by **4.5 hours**.
- **HPC Orchestration**: Developed an ML orchestration layer for distributed GPU training on HPC clusters. Engineered checkpoint-resume logic that handled preemptive node shutdowns, optimizing resource utilization and reducing compute waste by **15%**.
- **MLOps**: Reduced research environment setup time from hours to minutes by containerizing fine-tuned models with **Docker** and managing the experimental lifecycle (versions, hyperparameters, and weights) via **MLflow**.

## Software Engineer | UJR Technologies Pvt Ltd | Hyderabad, India

**Jul 2021 - Dec 2022**

- **API & SDK Development**: Designed and maintained standardized **REST APIs** and **Python-based SDKs** to streamline the ML development lifecycle, reducing cross-team integration defects by **40%**.
- **Model Serving**: Engineered model-serving endpoints with automated input validation and deployment health checks, lowering prediction-related failures by **30%** for ML-driven features.
- **CI/CD Pipeline**: Automated CI/CD pipelines via **GitHub Actions** with comprehensive test coverage and scripted rollback procedures, decreasing release failures by **20%** across production environments.