role: AI Engineer @ Infovibes
education: MS Computer Science, FAST National University (CGPA 3.88/4)
focus: LLM Systems · Agentic AI · NLP · Computer Vision · Healthcare AI
publication: TransFINN — IEEE Access, 2025
location: Karachi, PakistanI build AI systems that go from research paper to production — not just prototypes. Over 2+ years I've shipped LLM-powered automation pipelines, context-aware conversational agents, and applied deep learning systems across fintech, healthcare, and enterprise workflows.
What sets my work apart is the research foundation behind it: I'm a published IEEE Access author (2025), co-authoring TransFINN — a transparent, interpretable feature selection framework for text classification. That "explainability first" mindset shows up in how I build: not just models that perform, but systems people can trust and understand.
Currently: building AI-driven automation pipelines and conversational systems at Infovibes, integrating LLMs with external tools, databases, and APIs for end-to-end workflows.
🔭 What I work on:
- Agentic workflows & RAG pipelines (LangChain, FAISS, Hugging Face)
- Conversational AI with multi-step task handling
- Applied deep learning for healthcare & medical imaging
- Low-resource NLP (built an English→Sindhi NMT model from scratch)
💬 Ask me about: LLM system design, RAG architecture, interpretable ML, or my published research on transparent feature selection.
|
Languages
AI / ML & Deep Learning
Web & API
|
Tools & Platforms
Databases
Deployment & Cloud
Domains
|
|
Multi-class skin cancer detection mobile app pairing deep learning classification with an LLM layer that generates suggestions, precautions, and diagnosis support in plain language.
|
Secure, Azure-hosted assistant with a FAISS-backed RAG layer for accurate retrieval and a React TypeScript frontend, deployed via Vercel.
|
|
Transformer-based NMT model built from scratch — custom WordPiece tokenizers, encoder-decoder architecture, trained for a genuinely low-resource language pair.
|
Published research on transparent, interpretable feature selection for text classification, co-authored with Dr. Muhammad Rafi.
|
|
BERT-powered question-answering system trained on doctor-patient conversation transcripts, delivering context-aware answers through a user-friendly interface for patients and clinicians.
|
AI-powered Forex market prediction system using Claude AI to reason over structured economic calendar data.
|
- MS Computer Science — FAST National University, Karachi (2025–2027, CGPA 3.88/4)
- BS Computer Science — Sukkur IBA University (2019–2023, CGPA 3.3/4)
- 🏆 Sindh Talent Hunt Program · HEC Fully Funded Merit-cum-Need Scholarship · Runner-Up, Robotics Competition (Solar Panel Cleaning System)
I'm always open to conversations about LLM systems, applied NLP, or interpretable ML — reach out if any of that overlaps with what you're building.
