MLOps
PolicyOps
End-to-end MLOps pipeline for insurance fraud detection — F1 0.84, ROC-AUC 0.92, with SHAP-based explainable predictions.
Architecture, pipeline, and frontend
PythonXGBoostDVCMLflowSHAP
About
Background, engineering focus, and the principles driving my work in applied artificial intelligence.
Who am I?
I am an artificial intelligence engineering student pursuing my Bachelor of Science in AI at SZABIST Karachi. My work sits at the intersection of modern autonomous intelligence and scalable software systems, translating theoretical research into resilient production architecture.
What do I focus on?
My engineering practice focuses on designing multi-agent orchestration frameworks, production-grade retrieval-augmented generation (RAG) pipelines, and reproducible MLOps infrastructure. I build systems with strict attention to latency, continuous evaluation, and deterministic error handling.
Why this work?
I approach engineering as a craft grounded in clarity, discipline, and intention. For me, artificial intelligence is about building dependable, high-leverage systems that solve tangible problems and meaningfully expand what people can create and achieve.
Featured Projects
Case studies and builds across agentic AI, RAG pipelines, MLOps, and full-stack engineering.
Case-Study Projects
Agentic AI
ThesisLens
Autonomous research intelligence agent over ArXiv — sub-second Groq reasoning with interactive D3.js knowledge graphs.
LangGraph agent orchestration, FastAPI, and D3.js graph
LangGraphGroqArXiv APIFastAPID3.js
NLP / FinTech
MarketSynapse
Financial intelligence platform fusing FinBERT news sentiment with live yfinance price feeds into on-demand market briefs.
FinBERT sentiment pipeline, LangGraph agent, and frontend
FinBERTLangGraphyfinanceFastAPIReact
More Projects
Skills & Tooling
Core toolchain and production technologies across languages, agentic AI, MLOps, and full-stack systems.
Languages
Python
JavaScript
TypeScript
Java
C
C++
SQL
Frontend & Full-Stack
React
Next.js
Vite
Tailwind CSS
Three.js
D3.js
React Native
Expo
JWT
MLOps & Infra
MLflow
DVC
SHAP
FastAPI
Docker
Kubernetes
GitHub Actions
CUDA
Supabase
PostgreSQL
AI/ML & LLM Tooling
PyTorch
TensorFlow
Keras
scikit-learn
XGBoost
Transformers
OpenCV
NLTK
LangGraph
LangChain
HuggingFace
Groq
Gemini API
OpenAI API
RAG
ChromaDB
FAISS
MCP
Tools & Platforms
Git
GitHub
Streamlit
Vercel
Railway
Google Cloud
Google Colab
Firebase
Jupyter
Experience
Roles and hands-on work in AI/ML engineering.
Aga Khan University Hospital (AKUH)
AI/ML Intern
July – August 2025
Focused on medical imaging research using deep learning segmentation techniques.
- Implemented nnU-Net for pediatric MRI segmentation (LISA Challenge) — mean Dice 0.68, IoU 0.52
- Researched MRI super-resolution and segmentation architectures, including emerging MAMBA-based approaches
- Evaluated model outputs through rigorous quality control
Alphatron Technologies
AI Engineering Research Intern
2026 — Present
Conducting applied machine learning research and engineering across intelligent systems.
- Engineered end-to-end data science and reinforcement learning pipelines
- Designed and evaluated autonomous agentic workflows under iterative supervisor feedback cycles
- Architecting a multi-agent incident response and on-call triage system using LangGraph, ChromaDB, and MCP
Achievements
Competition placements and recognitions.
2ND
2nd Place — National AI Hackathon '26 (Karachi Regional)
atomcamp·May 2026
Built EduFlow, a 9-agent AI platform for Pakistan's education gap, competing against ~60 teams (~176 participants).
1ST
1st Prize — Machine Learning Competition
ZABEFEST (SZABIST)·2025
XGBoost pipeline predicting liver cirrhosis survival outcomes on the Mayo Clinic dataset — log-loss 0.3557, 1st place on the leaderboard.
3RD
3rd Prize — AI Image Generation Competition (World No Tobacco Day)
Community Health Sciences Department, Aga Khan University·2025
AI-generated public health awareness imagery for tobacco-use prevention.
Let's Work Together
Open to internships and collaborations in agentic AI and applied ML.
SEND A MESSAGE
Fill out the details below to open a pre-filled draft in your default email client.
DIRECT CONTACT
However you prefer to reach out — it's all below.