What I Do
I design and engineer AI systems — from data pipelines and deep learning models to LLM-powered products. I care about systems that actually ship.
AI / ML Engineer
What I Build
Agentic workflows with LangChain, RAG retrieval stacks, audio & vision classifiers in PyTorch, and full-stack interfaces around them.
The stack
behind the work.
AI/Machine Learning
- Python
- PyTorch
- ResNet
- TensorFlow
- Scikit-Learn
- ONNX
- NumPy
- Pandas
- NLP
LLMs & Agentic AI
- LangChain
- RAG
- Prompt Engineering
- Hugging Face
- Vector Databases
- Fine-tuning
- Quantization
Projects
VoiceGuard
Audio classification pipeline · PyTorch · ONNX
Problem
The rapid rise of highly realistic AI-generated synthetic media has made distinguishing between genuine human voices and deepfakes incredibly difficult, compromising biometric and communication security.
Approach
Engineered an audio feature-engineering pipeline that captures distinct acoustic anomalies, trained a robust PyTorch classifier optimized to withstand synthetic variance, and exported the final model to ONNX for lightweight, real-time threat detection.
- Python
- PyTorch
- ONNX
- NumPy
- Pandas
- Audio Processing
JobGuard
raudulent posting classification • Scikit-learn • XGBoost
Problem
The rise of sophisticated online job scams risks user data and financial security, yet detecting them requires parsing a complex mix of deceptive text and manipulated metadata.
Approach
Built an integrated data pipeline that extracts linguistic features alongside categorical indicators, resolved the heavy class imbalance using ensemble learning, and tuned classification thresholds to minimize false negatives.
- Python
- Scikit-learn
- XGBoost
- NLTK
- Pandas
- NumPy
What I’m
exploring now.
- In progress
Multi-Agent Systems
Coordinating specialised agents that plan, critique, and act together.
- In progress
Agentic AI
Tool-using LLM agents with memory, planning, and self-correction loops.
- In progress
MCP
Model Context Protocol — composable tool servers for AI clients.
- In progress
AI Product Engineering
Shipping AI features users trust: latency, evals, guardrails.
- In progress
RAG Architectures
Hybrid retrieval, reranking, and structured grounding for accuracy.
- In progress
AI Evaluation
Offline + online eval harnesses, golden sets, and behavioural tests.
From the
commit log.
A selection of repositories. The full archive lives on GitHub — including experiments, course work, and works-in-progress.
github.com/waniazanib ↗- repo / 01Python
Moodfit
Poetry-to-outfit multi-modal semantic search engine.
Open repository ↗ - repo / 02TypeScript
Insights
Instant product analysis and community sentiment retrieval
Open repository ↗ - repo / 03Python
Cold_Mails
Autonomous multi-agent pipeline that drafts personalized cold outreach emails
Open repository ↗ - repo / 04Python
Mask-CV
Video face-masking tool with identity-aware tracking
Open repository ↗
Resume.
The latest version of my resume — current education, projects, and the technical stack I work in day to day.
Let’s build
something.
Interested in AI, machine learning, agentic systems or automation? I’d love to hear from you — for collaborations, or just a good conversation.