Machine Learning Engineer Intern at Immersion Labs

Developed an end-to-end gesture recognition pipeline using TensorFlow, NumPy, and Leap Motion sensor streams; engineered features for 10K+ samples and trained deep neural networks achieving 99% accuracy with 30% faster training after optimization. Built a low-latency inference service using Docker, Redis Streams, and AWS EC2/GPU instances, delivering <50 ms real-time responses for AR/VR interactions. Created ML observability dashboards (Tableau + Python) to monitor accuracy drift, latency anomalies, and model performance trends, reducing debugging and engineering triage time by 34%. Implemented reproducible MLOps workflows using Docker, GitHub Actions, and AWS ECS/RDS for automated builds, tests, and deployments, reducing environmental inconsistencies by 50%.

AI & Data Engineer at USA National Phenology Network

Designed scalable event-driven data pipelines using Node.js, SQL, and AWS Lambda/S3 to capture and process 5K+ monthly interactions across Phenology & GeoServer portals, generating reliable ML-ready datasets and reducing manual preprocessing effort by ~25%. Developed internal AI assistants using GPT-4, Gemini, Groq, and Mistral to summarize research documentation and generate automated insight reports, reducing manual reporting effort by ~40%. Built Tableau dashboards with MySQL + AWS RDS integrations to visualize portal usage clusters and research tool performance, improving actionable insights by ~60% across the science team. Implemented early prototypes of RAG pipelines (embeddings, vector search, metadata indexing) enabling automated Q&A over historical ecological datasets for researchers. Built a LangChain-based workflow to auto-generate dataset summaries and species-level observation notes from raw phenology records, helping researchers quickly interpret data uploads and reducing