Ankit Sati

• Designed and deployed a firmwide AI-driven risk analysis utility, integrating JPMC’s structured risk inventory with ~500k biweekly news articles and LLM pipelines for real-time thematic detection and automated risk signal extraction. • Fine-tuned custom domain-specific transformer models (~20M parameters) on proprietary datasets for scenario classification and risk tagging, achieving ~92% precision across multi-LOB applications. • Developed a quantitative stress testing engine leveraging outputs from the risk platform and combining Vector Autoregression (VAR), Structural Equation Modeling (SEM), and XGBoost to simulate macroeconomic shocks and identify key scenario drivers. • Integrated GenAI models (GPT-4, BART, Pegasus) and BERTopic-based semantic clustering into the pipeline, enhancing unstructured data comprehension and elevating data quality to ~96%.

• Among the first 9 engineers in the Decision Sciences team, I helped design an anomaly detection system that improved hardware and software reliability by anticipating potential issues that could cause production impacts or sitewide outages. • Aggregated data from Director Switches and Storage devices across over ~800 features (~400GB) to deploy a pilot machine learning model, achieving an F1 score of ~0.72 and successfully identifying 2 Severity-1 production-impacting issues. • Collaborated with subject matter experts in Deep Learning and SAN infrastructure to fine-tune models for client-specific environments, improving prediction accuracy(~by 8%) and resolving the bottleneck issue in scale data centers. • Led the growth of our team from 9 to over ~100 engineers, transforming it into a cross-functional department. Trained and mentored members in ML, DL, SOTA models and optimization techniques to facilitate a seamless integration process. • Achieved zero production impacts on the DXC data center by predicting multiple(~14) Sev-1 and Sev-2 scenarios, making this model a key element in HPE's ~99% SLA and part of the Data Center Care contract. **Migration of Storage Ecosystem from HORCM(P9500) to Python3(XP7) – CI/CD, Linting, Docker, JavaScript, ReactJS** • Implemented automated tests using continuous integration CI/CD pipelines through Jenkins to ensure backward and forward compatibility across both Python versions. • Optimized the legacy codebase by leveraging modern Python 3.x features such as asyncio for asynchronous programming and utilizing faster Python 3 implementations. **SAN Framework Monitoring – Dashboards, SAN Deployment, Sandboxes** • Created Dashboards to monitor the SAN environments and optimized workflows for new use cases and devices. • Ensured 99% SLA while working on Disaster Recovery infrastructure in outage scenarios and reduced the downtime by 60%.

• Led the automation of test plans for various server components, such as BIOS, BMC, storage controllers, and networking interfaces. • Offered integral support through new installations, documented best practices, and actively contributed to hardware, software, drive, and interface installations, with continuous monitoring for seamless integration. • Worked in outage scenarios to to bring the production back online and provide clients a Root Cause Analysis. • Creating predictive analysis models with Brocade Network advisories to create data flow architecture. • Periodic checks/upgrades on EVA, XP, MSA and SDN’s for bugs to prevent any potential threat on the production.