Aafia Azhar
Computer Systems Engineer · AI/ML Engineer · Full Stack Developer
Computer Systems Engineering graduate with hands-on machine learning internship experience across NLP, computer vision, and predictive analytics. I build full-stack AI-powered applications, fine-tune LLMs, train vision models, and take ideas from research and development through to deployment.
Expertise
A focused toolkit for building, deploying, and evaluating AI-powered software.
Programming & Databases
Machine Learning
Deep Learning & Generative AI
Computer Vision & XAI
Web & Full-Stack
Tools & Deployment
Work Experience
Generative AI & Prompt Engineering Intern
- Practiced zero-shot vs. few-shot, chain-of-thought, and persona prompting; built a reusable prompt library, a custom chatbot, and a RAG mini-project for document-grounded Q&A with structured JSON output.
- Explored multi-agent systems and built a no-code automation workflow connecting an LLM to a real-world task pipeline.
Machine Learning Intern
- Built an AI content-marketing automation system (GPT + Unsplash API + Google Apps Script) that cut blog/caption creation time from hours to seconds.
- Developed a multilingual AI voice-cloning studio with noise-reduction filtering, and a 6-class emotion detection model combining TF-IDF, VADER, and XGBoost.
Machine Learning Intern
- Built regression models for student-score prediction and time-series sales forecasting (XGBoost), evaluated with MAE, RMSE, and R².
- Built a K-Means customer-segmentation model, a collaborative-filtering movie recommender (SVD, Precision@K), and a loan-approval classifier handling imbalance with SMOTE.
AI/ML Engineering Intern
- Built regression models for house-price and stock-price prediction, and reached strong accuracy on heart-disease risk classification.
- Fine-tuned BERT for 4-category news classification and built a multimodal CNN + tabular housing-price model alongside a production-style scikit-learn churn pipeline.
Education
B.E. in Computer Systems Engineering
PEC accredited · Level II · Washington Accord substantial equivalence.
Intermediate, Pre-Engineering
Percentage: 96.36% · Grade A1.
Undergraduate Thesis
The flagship project tying together the rest of the work below.
AutoGrade — AI-based self-assessment system using image-to-text conversion and LLM grading
Co-developed an end-to-end AI assessment platform that extracts handwritten exam answers using a multimodal LLM (Llama 4 Scout) and grades them semantically against rubric criteria, rather than relying on keyword matching. Fine-tuned a Llama 3.2-3B Instruct model with LoRA and 4-bit quantization for the grading engine, and integrated a full-stack pipeline — React, Express.js, PostgreSQL — around it, reaching strong text-extraction accuracy and close agreement with expert human grading.
Research Interests
Areas I’m actively exploring through projects, experimentation, and research.
Machine Learning & Deep Learning
Real-world applications of learning systems, predictive modeling, and intelligent automation.
Computer Vision & Generative AI
Vision models, multimodal systems, and generative AI for practical applications.
NLP & LLM Fine-Tuning
Natural language processing, transformers, prompt engineering, and parameter-efficient fine-tuning.
Physics-Informed ML
Combining machine learning with domain knowledge to model complex scientific systems.
Explainable AI (XAI)
Interpretable machine learning using techniques such as SHAP, Grad-CAM, and LIME.
Projects
Face Mask Detection System
Real-time detection using CNN (MobileNetV2) and OpenCV, served through a Flask dashboard for live classification.
Breast Cancer Classification System
XGBoost classifier trained on the Breast Cancer Wisconsin (Diagnostic) dataset, wrapped in a Flask web app with a JSON API and live accuracy/precision/recall/ROC-AUC reporting. Built for portfolio and learning purposes, not clinical use.
AI Content Marketing Platform
Automated content generation and analytics tracking built on Node.js and Express, wired to LLM APIs to draft and schedule marketing copy.
SafeRoute — Smart Safety App for Women
One-tap safety app in Kotlin with instant SOS alerts, live location sharing, and offline SMS fallback for emergencies.
AI Resume Analyzer & Interview Prep
Full-stack ASP.NET Core 8 MVC app with resume parsing, ATS scoring, skill-gap analysis, and an interview-prep module with an analytics dashboard.
Lexicon & Transformer-Based Emotion Detection
Multiclass emotion classifier combining lexicon methods with Transformer models to analyze written text with high accuracy.
Satellite Flood Detection from Imagery
From-scratch U-Net benchmarked against a pre-trained EfficientNetB4 model for flood segmentation on satellite imagery in TensorFlow.
Smart Traffic Controller, Emergency Priority
Digital controller designed in Verilog HDL with FSM-based real-time routing that overrides normal signal flow for emergency vehicles.
Taj Mahal — 3D CAD Model
Full 3D AutoCAD model of the Taj Mahal built with and without marble surface textures, with reference photos used to match real proportions and materials.
Smoke Detection & Fire Prevention System
Simulated IoT smart-home network in Cisco Packet Tracer where smoke detectors trigger a central gateway to coordinate doors, windows, garage, sprinklers, and an alarm over DHCP-assigned devices.
Certificates
15 certifications across ML/AI, volunteering, and workshops. Filter by category, tap any to view.
Get in Touch
I welcome opportunities for collaboration, research engagement, and conversations on how AI and machine learning can address meaningful real-world challenges.