Hi, I'm Neha Malviya

Patient satisfaction was declining at Capital Health's busiest campus (RMC), with billing complaints and coverage gaps affecting an estimated 872,000 NJ residents.

Cleaned and analyzed four real hospital datasets provided directly by Capital Health, spanning a full year, built per-campus ARIMA admission forecasts, and ran sentiment analysis on patient reviews across all three campuses.

Traced RMC's −0.30 composite sentiment score to sustained peak-hour overload (9 AM–6 PM) rather than clinical failure, and delivered six data-backed operational recommendations.

40.6% of Fulton's 38,644 Gen Z household customers had zero mobile app logins in one month — Fulton risks losing this generation to fintechs like Chime and Cash App.

Worked directly with Fulton Bank SMEs under a tight timeline, ran our own survey research alongside 453 App Store/Google Play reviews, and analyzed Fulton's internal Gen Z household data and competitor products to uncover engagement gaps and build six prioritized recommendations plus two new product concepts.

Delivered a 12–18 month roadmap with quarterly KPIs, a dormant-user re-engagement campaign, and two original loyalty programs tied to real spending data and revenue projections.

Business teams waste hours manually preparing sales reports and meeting summaries from scattered data.

Streamlit dashboard that auto-detects columns, generates trend charts, and emails PDF reports with one click.

Meeting prep time cut from hours to minutes. Automated week-over-week insights delivered to stakeholders.

Companies receive thousands of unstructured reviews making it impossible to manually extract trends and sentiment.

Universal NLP pipeline: spaCy lemmatization → TF-IDF vectorization → K-Means clustering → TextBlob sentiment scoring.

Processes up to 3,000 reviews per run. Groups into up to 10 topic clusters with per-cluster sentiment breakdown.

Over 135,000 Olympic athlete records spanning 51 games are complex and hard to explore without interactive tools.

Dynamic filtering webapp for medal tallies, heatmaps, athlete distributions, and men-vs-women trends.

Four analysis views: Medal Tally, Overall Stats, Country-wise, and Athlete-wise — all interactive.

False-positive cancer screenings lead to costly, unnecessary surgeries. Doctors need reliable ML decision support.

Flask web app with SVM classifier trained on 569 patient records. Both ANN and SVM evaluated — SVM won.

98.2% accuracy. Real-time predictions with confidence scores reducing unnecessary surgical interventions.

Campus events are scattered across emails, Instagram, and flyers — students miss events they'd love, and departments see low attendance.

Centralized React web app with category filtering, @rider.edu email verification, admin approval workflow, and organizer status tracking.

Live at rider-eventhub.netlify.app. One hub for all campus events — searchable, filterable, mobile-friendly, and admin-controlled.

Explore how Claude AI can be embedded into a Streamlit app to create a responsive, intelligent chat experience.

Built a lightweight chat interface in Python using the Anthropic SDK, with conversation history, styled message bubbles, and real-time AI responses.

A working demo that proves how quickly a functional AI chat interface can be built end-to-end using modern generative AI tools.

Marketing and business teams struggle to monitor real-time web traffic, user behavior, and engagement metrics in one place.

Power BI dashboard connected via Google Analytics API with complex DAX measures for engagement rates, session trends, and user conversion patterns.

95%+ data accuracy. Live KPIs including traffic sources, bounce rates, and geographic performance — updated in real time.

Olympics data is vast and scattered — hard to explore trends, compare countries, and surface athlete-level insights in one place.

Built an interactive Streamlit dashboard with dynamic filtering by year and country, visual heatmaps, and athlete-level breakdowns.

Rich visual storytelling of Olympics history — from overall participation trends to individual athlete performance across sports and events.

HR leadership had no self-service way to explore workforce data — every question meant a manual pull and a one-off report.

Designed a SQLite schema from scratch, authored 5 SQL queries tracking core workforce KPIs, and deployed a live Streamlit dashboard with 8+ interactive visualizations and sidebar filters.

Detected that overtime-mandatory teams had 2× the attrition rate of other teams — surfaced through multi-factor SQL pattern analysis and packaged as an executive summary with trend visualizations.