Aditya Prabhudessai
Bridging the gap between
Data Science and Interactive Web Experiences.
About
A final-year B.Tech student at Manipal University Jaipur, interested in solving real problems through Data Analytics, Machine Learning, and Full-Stack Development.
I value clarity, thoughtful design, and solutions that scale beyond demos.
Technical Skills
Python
TypeScript
JavaScript
C++
Java
PyTorch
Scikit-learn
LangChain
NLP
Pandas
NumPy
Seaborn
Matplotlib
Power BI
Tableau
MySQL
MongoDB
React
Next.js
Tailwind
Node.js
FastAPI
AWS
GCP
Git
GitHub Actions
Experience
- Built Jot It, an AI-powered full-stack note-taking app with handwriting OCR via Azure Computer Vision, returning per-word confidence scores.
- Architected a three-service system: ASP.NET Core 8 REST API, Python FastAPI OCR microservice, and a React Native mobile app (Expo SDK 54).
- Implemented JWT authentication with BCrypt hashing, full CRUD note management, and paginated endpoints backed by PostgreSQL.
- Deployed as a PWA with offline caching, dark/light theme, and a canvas-based drawing surface; containerised OCR service with Docker on Render.
- Built and evaluated forecasting models on 1M+ time-series records, improving prediction accuracy by 15%.
- Performed data preprocessing, anomaly detection, and missing-value handling using resampling techniques.
- Engineered time-based and statistical features to enhance model stability.
- Optimized ML pipelines to reduce training and experimentation time by 20%.
- Developed end-to-end ML pipelines with cross-validation, improving robustness by 12%.
- Assisted in building LLM monitoring dashboards to track drift, latency, and quality.
- Conducted error analysis and documented findings to support model deployment.
- Automated ETL workflows and BI dashboards, reducing manual reporting effort by 40%.
- Analyzed datasets with 500k+ rows to identify data quality gaps.
- Created structured reports to communicate insights to non-technical stakeholders.
Featured Projects
Full-stack AI note-taking app with handwriting OCR. Draw on a canvas and Azure Computer Vision extracts your handwriting with per-word confidence scores. Three-service architecture with JWT auth, React Native mobile app, and an installable PWA.
Domain-specific AI consultant for a structural engineering firm covering property valuations, RCC design, structural audits, and legal advisory. RAG pipeline over 100+ Q&A pairs; automated lead capture to Google Sheets with email alerts.
Analyzes YouTube comments to determine sentiment and user personas using BERT. visualizing public opinion trends effectively.
LSTM-based text classification model with SHAP and LIME for explainability. Analyzed misclassified samples to reduce model bias.