Jul – Aug 2026 · Masar Internship Project
Flydubai AI Support System – RAG Email Support System
Completed
Self-initiated during the Masar Internship starting in
Rotation 01 (IT Ops/Cargo/CX) and submitted upon internship completion: Gmail API ingestion, an HTTP MCP
server standardising the agent's tool calls, Gemini 3.1 Flash classification and reply drafting grounded
in FAISS-retrieved case history, a humanizer pass, and human sign-off before send.
HTTP MCP Server
Gemini 3.1 Flash
FAISS Grounded
Human Sign-off
Jun 2026
The Human Skill - Claude Style Transfer Skill
Completed
Custom Claude skill that detects and strips AI writing patterns from text — then
rewrites output in a neutral human register or user's documented style. Structured prompt harness analyzing
vocabulary & sentence distribution.
Style Transfer
SKILL.md Arch
Claude APIPrompt Eng
Style TransferSKILL.md
Mar – May 2026
AutoGPT Skill Gap Analyser
Completed
Custom AutoGPT agent blocks chaining LLM tasks — resume parsing, job description
matching, and gap scoring — into an automated workflow. Part of 3-person team; owned frontend UI & block
orchestration.
AutoGPT blocks
LLM Chaining
PythonAutoGPT
DockerReact
Tailwind CSS
Sep – Dec 2025
Churn Prediction Pipeline (Spotify)
Completed
Reusable preprocessing pipeline (IQR outliers, engagement ratio, StandardScaler).
4-layer TF/Keras MLP (256→128→64→32→1) with SMOTE, BatchNorm, Dropout, EarlyStopping; 93% accuracy &
business ROI analysis.
93% accuracy
Business ROI
TensorFlowKeras
Scikit-learnSMOTE
Seaborn
Jul – Oct 2025
Real-Time Face Recognition System
Completed
3-stage iteration: Haar cascades → dlib ResNet 128-d encodings → TF/Keras age &
gender classification. Real-time frame optimization (every N frames caching), live confidence overlays, and
pickle-persisted face database.
Multi-person RT
ResNet 128-d
OpenCVface_recognition
TensorFlowDlib
Caffe DNN
Jul – Aug 2025 · Internship
Customer Churn Prediction - Telecom ML Model
Assessed
End-to-end pipeline on telecom dataset: EDA, feature engineering, dummy encoding,
StandardScaler, and Logistic Regression. 88% accuracy on both train and test sets, validated via confusion
matrix.
88% accuracy
NSDC assessed
PythonScikit-learn
Logistic RegressionEDA
Jun – Jul 2025 · Internship
Sentiment Analysis - Movie Reviews
Assessed
Binary sentiment classifier using TF-IDF vectorisation and Logistic Regression on
movie-review benchmark. Reached 85% test accuracy after iterating on preprocessing and feature weighting.
85% test accuracy
Skill India
PythonNLTK
TF-IDFLogistic Regression
Mar – Aug 2025
CB Triage - Healthcare Web & Mobile App
Completed
3-phase web app (lab results, symptom tracker, appointments) converted to cross-platform
Flutter mobile. Client-side UI built solo; Firebase Auth & Cloud Firestore scoped for backend
integration.
Web → Mobile
Firebase
Google AI StudioFlutter
DartFirebase
Sep 2024 – Mar 2025
Driver Drowsiness Detection System
Completed
Driver safety prototype monitoring real-time eye aspect ratios via Dlib facial
landmarks, triggering an audio alert upon fatigue detection.
Real-time EAR
PythonOpenCV
Dlib