Kiranmai

Data & Operations Analyst — ML Systems & AI - Uber - Hyderabad, India

(2023-05 - 2025-11)

Worked on large-scale ride-booking, dispatch and driver-allocation data feeding Uber's ML-driven dispatch and ETA prediction systems — investigating anomaly patterns, model output deviations and data drift in production.
Built validation and pattern-detection logic in Python and SQL across geo-location, time-of-day and demand signals — feature-engineering-adjacent work that surfaced quality issues in upstream training and inference data.
Performed root cause analysis on production KPI deviations (booking success, ETA accuracy, error rates) and partnered with engineering and product teams to ship fixes.
Automated repetitive validation pipelines using Python, cutting manual QA effort on critical ML-adjacent workflows.
Explored LLM-based summarisation for internal incident reports using Python and the OpenAI API — built a prototype that auto-summarised long RCA documents into structured formats, reducing review time for the ops team.
Studied prompt engineering and Generative AI concepts as part of ongoing upskilling — applying learnings to automate repetitive text-heavy reporting tasks within the team.
Operated in an ambiguous, fast-moving environment — picked up unfamiliar tools, scoped own investigations and shipped without waiting for handed-down tickets.

Ward Education & Data Processing Secretary - Govt. of Andhra Pradesh - Andhra Pradesh, India

(2019-10 - 2023-09)

Owned end-to-end data operations for 800+ students across 12 schools — built structured reporting workflows using Python and Excel, going from paper registers to clean datasets to district-level MIS submissions.
Prepared monthly MIS reports for the District Education Officer covering attendance trends, dropout rates, welfare scheme utilisation and demographic breakdowns — automated report generation using Python (pandas), adopted as standard inputs to district review meetings.
Analysed scheme enrollment and attendance data using Python to identify underutilisation patterns across student groups; flagged gaps that led to targeted outreach in 3 villages and measurable improvement in enrollment.
Designed validation, reconciliation and cross-check logic in Python for data collected across 6 government welfare programs, ensuring accurate and audit-ready submissions for district reporting cycles.
Sole data point of contact for the District Education Officer — figured out reporting workflows independently, coordinated with 12 schools and resolved escalations without a playbook.

About