Daniyal Khan

Selected Work

Things I've built

Each of these is written up end to end — the problem, the approach, what the numbers came out at, and what I'd do differently.

Voice Fingerprinting Pipeline

Isolating individual customer conversations inside eight-hour, multi-speaker audio recorded on wearables across retail stores.

~60%
Precision
~49%
Recall
100k+ hrs
Audio Processed
  • Python
  • PyTorch
  • Silero VAD
  • Pyannote
  • ReDimNet
  • Qdrant
  • Resemble AI
Read case study →

RAG-Based Event Detection Pipeline

Finding where one customer interaction ends and the next begins, by classifying what happened rather than who was speaking.

+59%
Customer Coverage
66%
Personal Speech Cut
2,000+
Stores Live
  • Python
  • Gemini
  • Qwen Embeddings
  • RAPTOR
  • HyDE
  • KNN
  • Qdrant
Read case study →

Diabetic Retinopathy Diagnosis with Interpretable CNNs

Screening retinal images for diabetic retinopathy severity, with visual explanations a clinician can actually audit.

0.82
Quadratic Kappa
5
Severity Classes
Grad-CAM + LIME
Explainability
  • Python
  • TensorFlow
  • InceptionV3
  • Grad-CAM
  • LIME
  • OpenCV
Read case study →

Multimodal Sentiment Analysis with Tensor Fusion

Reading sentiment from speech and text together, because tone routinely contradicts the words.

Audio + Text
Modalities
Tensor Fusion
Fusion
Transformer
Architecture
  • Python
  • PyTorch
  • Transformers
  • Librosa
  • melSpectrogram
Read case study →

Hybrid Movie Recommender with a Tracked MLOps Pipeline

A content-plus-collaborative recommender built as a reproducible, versioned pipeline rather than a notebook.

Hybrid
Approach
DVC + MLflow
Pipeline
AWS EC2
Deployment
  • Python
  • scikit-learn
  • DVC
  • MLflow
  • Docker
  • FastAPI
  • AWS EC2
Read case study →