Rithik Marudappa

A clinic-grade record system that turns discharge summaries, lab reports and prescriptions into a structured, scored, longitudinal patient record.

• Built LangGraph pipelines for clinical fact extraction, normalization and reconciliation across a FastAPI backend, inference service, clinician dashboard and React Native app; achieved 94% recall at 98% precision on held-out clinical notes, with evals focused on end-to-end clinical accuracy.
• Top features — Dynamic Insights, a personalized insight generated each morning where the model drafts across permutations and code ranks and verifies them; and a voice-based intake system built on Jev and GPT-Live-1.
• Exploring LSM-2-based wearable representation learning with an HbA1c prediction head for chronic disease management, starting with diabetes.

Sole engineer on the Tabulation Engine — product, architecture, demos, end to end.

• Built a compute engine (DuckDB) that AI agents configure to turn raw survey data into statistical crosstabs — significance testing, weighting, confidence intervals — matching output from industry tools like WinCross & Q
• Led RapidFlow — describe a survey in chat, and agents build and deploy it in an isolated sandbox. Replit/Lovable, for surveys
• Architected a multi-agent data processing pipeline; cut its latency 40% via caching & parallelization

SAP Analytics Cloud OEM — analytics tooling that serves complex query results to non-technical business users. Started as a Software Developer Intern (2023) before joining full-time.

• Drove the HANA Cloud migration of an integration service from proof-of-concept to production
• Implemented async batch processing for parallel analytics calls — ~18% faster page loads on large datasets
• Built a D3 team-analytics dashboard introducing a quantitative index for cross-team progress