
AI Intern
Prisalis Private India Limited · Aug 2025 – Dec 2025
- Built an end-to-end NLP pipeline to normalize and classify unstructured clinical chief complaints into standardized medical categories, improving downstream data usability.
- Designed a hybrid rule-based + machine learning system (regex, medical dictionaries, ML fallback) to handle spelling errors, abbreviations, negations, and multi-symptom clinical inputs.
- Improved automated labeling coverage from ~80% to 90%+ by expanding rule patterns, medical vocabularies, and normalization mappings through systematic error analysis.
- Implemented abbreviation expansion, spelling correction (SymSpell), fuzzy matching (RapidFuzz), and negation detection, significantly reducing noise in raw clinical text.
- Trained and evaluated TF-IDF-based classifiers (Naive Bayes, Logistic Regression), iteratively optimizing preprocessing to improve classification accuracy and consistency.
- Manual review effort by automating complaint normalization and categorization, enabling scalable

