Door categorieën bladeren
Ontdekken
Fiverr Pro
Nederlands
$
USD
I will develop custom deep learning and nlp models
Data Scientist, AI, Machine Learning, Deep Learning, Data Analyst
I am a data science enthusiast, certified by Google as Data Analytics Professional as well as John's Hopkins Certified Data Scientist. I like having a challenge and I like to solve it. My expertise in...
Over deze dienst
Want recommendations that actually reflect how your users behave not a static "people who bought X" table?
I build deep-learning recommendation and sequence models in Python: systems that learn from the order of user behavior to predict what someone will want next, the way production systems at Netflix and TikTok do.
What I build: Sequential recommenders (SASRec / GRU4Rec / transformer-based) Embedding & collaborative-filtering models NLP and sequence models for text and behavioral data Proper evaluation leave-one-out, Hit@K, NDCG, coverage & diversity (not inflated random-split metrics) An interactive demo app so you can see recommendations live
Recent work: I built a sequential movie recommender on the MovieLens 25M dataset implementing three models (Prod2Vec, GRU4Rec, SASRec) with a live demo that visualizes which past items drove each recommendation via attention weights. I've also shipped an end-to-end customer-analytics system that identified £403K of at-risk revenue for a retailer so I build models that tie to real business outcomes, not just leaderboard scores.
Tools: Python, PyTorch/TensorFlow, scikit-learn, pandas, Streamlit
Programmeertaal:
Python
•
R
•
MATLAB
•
SQL
API's:
Microsoft Computer Vision AI
Tools:
Jupyter-notitieboek
•
opencv
•
tensorflow
•
Excel
•
Colab
Frameworks:
Scikit-learn
•
keras
•
PyTorch
•
Panda
•
tensorflow
Veelgestelde vragen
What data do you need?
User-item interaction history (who interacted with what, and ideally when). I'll tell you if it's enough to model.
How do you measure success?
Leave-one-out evaluation with Hit@K and NDCG; honest metrics that reflect real next-item prediction, not leaked random splits.
Can I see it working?
Yes. Standard and Premium include a demo so you can test recommendations interactively
What if my data's too sparse for deep learning?
I'll tell you upfront and recommend a simpler, more reliable approach instead
