Deeksha Garg

• As a Site Reliability Engineer (SRE), I automated processes to enhance the incident management lifecycle, reducing MTTD (Mean Time to Detect), MTTT (Mean Time to Triage), and MTTR (Mean Time to Resolve) through AI-based solutions like early incident prediction, similar incident detection, change correlation, and automated RCA prefilling.
• Mentored team members on adopting AI-driven solutions, fostering innovation, and skill development across the SRE team.

AICopilot for Incident Management:

• Developed a GPT/Gemini-powered AICopilot, achieving a 90% satisfaction rate by the operations team. Features include:
• Incident Journal and Scribe Document Generation: Automates the creation and regular updates of incident summary documents and real-time scribe timelines for bridge discussions.
• Internal/External Communication Generator: Composes communication tailored to executives and external vendors, integrated with SLA reminders for ongoing incidents.
• Question Answering and Actionable Tool: Enables real-time responses and actionable tasks via Slack incident channels, private conversations, and FAQs, using RAG search techniques.
• LLM Response Evaluation Framework: Validates AI-generated responses against metrics like context relevance, accuracy, toxicity, and hallucination, using tools like Deepeval, Langchain, and Uptrain.

Merchant and Incident Analytics:

• Developed text-to-SQL query tools for dynamic visualization of merchant and incident data from BigQuery to support SRE decisions.
• Utilized MCP tools: automated workflows to fetch merchant, failed customer interaction (FCI), and change risk score data for improved visibility and response accuracy during incidents.
• Designed advanced dashboards to provide actionable insights for incident resolution and risk analysis.

Incident SRE Data Lake:

• Architected scalable data lakes to centralize data from ServiceNow, postmortem reports, merchant records, and AI insights.
• Automated Apache Airflow DAGs for real-time data synchronization, with seamless integration into Milvius Vector DB, to enhance search and incident intelligence capabilities.

Developer Productivity Enhancement (DPE) Workflow Automation.

• Created an Agentic AI using the Langgraph and CrewAI frameworks to automate Slack workflows, including ticket creation and resolution, without human intervention.
• Utilized reflection and React patterns to automate query handling, follow-ups, and resolution workflows, boosting developer productivity and velocity.

• Leading a team of 6 members working on a Sales Optimization Project.
• Working on several use cases related to the entire customer journey over an e-commerce platform.
• Category Management Dashboard: - Created Data Streaming Pipeline into GCP's BigQuery to transform raw Clickstream data and Purchase Orders data from different zones into usable format to display into Dashboards.
• Developing Dashboard using Apache Superset to display the different KPIs that are important for E-commerce and Retail like Visit KPI , Bounce Rate , Cart Abandonment Rate, Revenue Generation etc
• Building Model for Anomaly Detection and Root Cause Analysis based on different factors impacting the no. of orders and Sales using GCP's BigQueryML
• Forecasting the Demand of different products of different categories and types for upcoming weeks based on historical demand to help Category Managers in inventory planning.

• Understand the business problem, identify the key challenges, formulate the machine learning problem and provide/prototype solutions.
• Working closely with Devops team for model and dashboard deployments on cloud using docker and Kubernetes services of Azure and GCP.
• Tire Mileage Prediction: -Build regression model to predict the performance of tires based on certain parameters like pressure, temperature, tension, noise, wear, rolling resistance etc.
• French to English Translation Tool: A GUI tool that converts several documents pdfs, images from French to English language using image segmentation, transfer learning and OCR techniques
• AI Extrusion Tooling: Deep learning-based model that predict Die geometric shape (metal plate for tire tread design) from Tire Cut in order to reduce number of die test in plant to save cost in manufacturing. Model reduced the number of die test from 3 to 1.
• Digital DUS: Designed Data Streaming Pipelines for multiple dashboards to consolidate, clean, preprocess and transform the data coming through different regions using GCP's BigQuery to analysis the historical performance of tires based on certain parameters on worldwide level using Power Bi, Kibana and Elasticsearch and python framework like dash and plotly.
• SmartMetro Analytics: Building wear prediction model for metro trains where we have the time series data of various performance parameters like temperature, pressure, mileage, speed etc data from sensor mounted on each car of train in mat file, pre-processing of data, Exploratory Data Analysis and wear prediction model for metro train tires.

• Used Python to scrape, clean, analyse large dataset, and Identify patterns and trends in data sets using statistical methods.
• The Customer Life Value model concentrates on customer purchasing behavior, activity, services utilized, and average customer value, and deploys the model using a Flask API.
• Worked in an Agile Development Environment while supporting requirement changes and clarifications.
• Writing and debugging shell scripts. PL/SQL is used for querying.
• Created a detailed design document, unit test plans, and well-documented code in Java for new feature development.
• Deploy and test the code on UAT and production servers.

• Collaborated with distributed SQL and Java teams to develop robust database solutions.
• Executed comprehensive SQL development, including stored procedures, views, and normalization.
• Performed ETL operations to optimize storage and processing of large datasets.
• Utilized Spark to accelerate data warehousing operations.