Sagnik Ghosal Email & Phone Number

Seattle, WA, US

Seattle, WA, US

• Created global investigation dashboards with consolidated metrics, that act as a single point of reference, and save ~100 on-call operator hours annually.
• Developed integration tests and implemented canaries to automate API testing and detect issues before getting escalated to Sev tickets.
• Enabled prism feature for the DevOps Guru console, that allows simultaneous console access for up to 10 AWS accounts in one browser.
• Upgraded team pipelines from CDK v1 to v2, reducing yearly Sev-2 tickets by 5%.

Seattle, WA, US

• Developed an ML framework utilizing participant survey responses, real-time prevalence, and sliding windows to identify a sub-population for COVID testing in resource-constrained environments.
• Designed and implemented an algorithm to match specific assay/workflow requirements and collect real-time fluorescence data.
• Developed a DeDx framework in Python and R, using a two-step scikit-learn pipeline for improved diagnostic tool accuracy. Evaluated the system's performance to maximize predictive values and population coverage.

Seattle, WA, US

• Streamlined load testing with a One-touch mechanism, simplifying inputs, enabling flexible configurations, and automating test validity checks and metrics.
• Implemented changes in frontend and backend to send client-side 400 errors to deny listed customers with a custom error message asking to reach out to support.

Paris, France, FR

• Translated SAS code to Pandas/PySpark, reducing computation time from hours to minutes.
• Created PowerBI dashboards for clear data visualization.
• Collaborated with clients to resolve software-related technical problems and tackle new challenges.

Kolkata, West Bengal, IN

• Developed an unsupervised streaming anomaly detection framework with dynamic graphs and clustering, surpassing SOTA models on 31 real and 22 synthetic datasets in accuracy, precision, recall, F-score, and MCC.
• Framework parameters include overestimated k for KNN and model retraining frequency, presented in a research paper currently under review.
• Achieved 90% accuracy in a face recognition app using PCA and the ORL face dataset, and also wrote comprehensive reports on Multimodal Machine Learning and Subspace Learning.

IN

• Developed a fuzzy ensemble deep neural model to identify human activities using sensor multimodal data.
• The proposed technique improved SOTA accuracy by 5% and reduced computation time by 10%.
• Achieved better performance than SOTA models on four real-life human activity recognition datasets.
• Published the research findings in the IEEE Internet of Things Journal.

Raipur, Chhattisgarh, IN

• Led a team of 4 to develop a smart blood glucose and diabetes sensing model which achieved 10.7% MAE, 1.12 mg/dL bias, 91.89% accuracy, 94% sensitivity, and 82% specificity. Published the findings in IEEE Sensors.
• Developed blood hemoglobin level and anemia-polycythemia screening model with ±0.33 g/dL accuracy, 0.2 g/dL bias, and 90% sensitivity, with adaptive K selection in KNN and automatic brightness adjustment. Published the.
• Developed a smartphone-based model for measuring blood hemoglobin levels with an accuracy of ±0.32 g/dL and screening for anemia with 89% accuracy, and published the findings in IEEE Sensors Journal.

Bangalore, Karnataka, IN

• Contributed 1000+ MCQs along with their solutions on Network Analysis, Electrical Measurements and Measuring Instruments, Signals and systems.

Master Of Science - Ms, Data Science

Bachelor Of Engineering - Be, Electrical Engineering