Ammara Tahir

AI/ML Engineer specialising in NLP, LLM systems, Computer Vision, and Generative AI, with hands-on production experience building RAG pipelines, multi-agent LLMs, CNN classifiers, OCR systems, and voice AI end-to-end.

Delivered measurable impact: 94% classification accuracy, 90% reduction in manual data entry, and ~30% improvement in inference throughput across multiple internships. Proficient in Python, TensorFlow, scikit-learn, Flask, FastAPI, and LangGraph — driven by building intelligent systems that solve real problems.

AI / ML Engineer Intern - HEX Technologies - Remote

(2026-04 - 2026-05)

• Designed and deployed end-to-end AI applications using Python, Flask, and scikit-learn; improved inference throughput by ~30% through optimised preprocessing pipelines.
• Managed full ML lifecycle — data preparation, model training, evaluation, and production integration — with versioned codebases reducing integration errors.

AI Research & Development Trainee - Samsung Innovation Campus × Bramerz - Pakistan

(2026-02 - 2026-05)

• Selected for Samsung's globally recognised AI/ML program (2nd Pakistan cohort); trained in Python, Big Data, Machine Learning, and Deep Learning.
• Architected HumSabaq AI — production LLM platform with RAG Document Q&A, 4-agent SmartVoice pipeline (Whisper ASR → LLaMA 3.2 + TTS), and AI Timetable Manager with MongoDB Atlas session persistence.

Machine Learning Engineer - Arch Technologies - Remote

(2026-03 - 2026-04)

• Achieved 94% accuracy on Email Spam Classifier (TF-IDF + bigrams, Naive Bayes / LR / SVM) and built MNIST Digit Recogniser (MLP, RF, KNN on 70K samples) — both Flask-deployed with live metrics dashboards.
• Owned full pipeline: EDA, feature engineering, hyperparameter tuning, evaluation, and versioned GitHub deployment.

Artificial Intelligence Engineer - Afynix Digital & Elevvo Pathways - Remote

(2026-01 - 2026-03)

• Delivered Churn Prediction, Real Estate Price Prediction (Streamlit), and K-Means Customer Segmentation — full pipeline across EDA, SMOTE, training, and UI integration.
• Built XGBoost models, collaborative filtering (MovieLens), Walmart time-series forecasting, CNN traffic-sign classifier (OpenCV), and MFCC-CNN audio classification (GTZAN).

B.Sc. Computer Science - Computer Science - Riphah International University (2026-01)