PromptPilot
PromptPilot is an AI-powered assistant designed to help users confidently articulate their thoughts in interviews, presentations, and professional conversations.
Projects
GitHub repositories
A curated set of repositories from my GitHub profile with direct links and short descriptions of what each project does.
Sorani Kurdish Dialect Identification
Research code for fine-grained Sorani Kurdish dialect identification that integrates rule-based morphological features with transformer-based neural models.
AjamiMorph
Official implementation of AjamiMorph, a multi-method consensus framework for unsupervised morphological discovery in Hausa Ajami, including morpheme inventory workflows.
Abjad Orthographic Variation Benchmark
Benchmark and evaluation code for measuring LLM robustness under realistic orthographic variation, including diacritics removal, romanization, and code-switching in Arabic-script languages.
Academic Outreach Assistant
AI-powered assistant for automating academic outreach, generating personalized emails, managing collaboration pipelines, and tracking researcher engagement.
CADET: Combined Attention and Deep Autoencoder with GloVe Features
CADET is a research-grade toolkit for transformer-centric automated essay scoring with prompt-aware pipelines, neural regressors, and reproducible evaluation routines.
NeuroAdaptive: Affective Decoding from EEG
EEG-based valence-arousal recognition using spectral biomarkers and CNN-Transformer models with frequency-aware attention, designed to be reproducible and scriptable.
HashSeqLab
Reproducible research pipeline for labelled DNA sequence classification with hashed k-mer features, distributional representations, and structured experimentation.
VisionAid Assistive Suite
Dual Streamlit application that supports blind and low-vision users with Indian currency identification and nearby obstacle detection using depth-aware pipelines.
Aurora: Sight to Sound
Research pipeline that converts images and videos into musically coherent compositions by mapping visual semantics, color, and texture into interpretable musical attributes.
Resume Insight Engine
Modular framework for parsing resumes, engineering features, scoring candidate fit, and surfacing explainable insights for practical hiring intelligence workflows.
ShopSense Search
Interactive semantic product search system with transformer reranking and attribute-aware refinement to improve practical e-commerce retrieval quality.
Adaptive Guided Query Retrieval (AGQ-IR)
Conversational information retrieval engine that fuses semantic embeddings with structured attributes and iteratively refines search for stronger relevance.
Courses at SKEPSIS
Teaching sessions and workshops
During my time at SKEPSIS, I taught and mentored through a sequence of practical AI and ML sessions designed to help learners move from fundamentals to implementation. The structure balanced conceptual depth with hands-on coding, so students could understand not just what to build, but why specific modeling decisions matter. These sessions focused on clarity, experimentation discipline, reproducibility, and confident technical communication.
Course Session 1
An applied introduction to machine learning pipelines, from data preparation and feature design to evaluation strategy and model interpretation for project readiness.
Course Session 2
A deeper session on neural modeling choices, training behavior, debugging patterns, and practical optimization habits for better model performance and stability.
Course Session 3
A project-focused workshop on turning ideas into usable prototypes, communicating technical trade-offs, and developing collaboration habits for research and product teams.
If you made it here, we would probably have a great conversation about ambitious ideas and how to build them properly.