Machine Learning Intern
Cognifyz Technologies
Hyderabad
09.2025 - 11.2025
- Engineered a multi-model NLP classifier on a 24K-record Hate Speech corpus, benchmarking Logistic Regression, Linear SVM, and DistilBERT to identify the optimal architecture for production.
- Improved baseline F1 by +12 points through TF-IDF feature engineering, class-weight balancing, and grid-search hyperparameter tuning, reaching 94% accuracy on the held-out test set.
- Designed reusable preprocessing pipeline (tokenization, lemmatization, stop-word filtering) reducing inference latency by 35% versus the unoptimized baseline.
- Built an AI agent that ingests syllabus PDFs and auto-generates priority-weighted study schedules, processing 200+ pages in under 90 seconds.
- Integrated Llama 3.1 with structured prompting to extract topics, deadlines, and weightings at 96% extraction accuracy across 15 sample syllabi.
- Shipped automated email-reminder and PDF-export modules, cutting manual study-planning time by an estimated 4 hrs/week per user.
- Developed and deployed machine learning models in Python and TensorFlow for restaurant-rating prediction and cuisine classification, achieving 98% accuracy on the production validation set.
- Collaborated with a team of 5 engineers in iterative testing cycles, reducing model error rate by 22% across four sprint releases.
- Authored evaluation dashboards tracking precision, recall, and F1, accelerating model-comparison decisions by 3×.