Hybrid Recommendation Tech Lead Needed
Budget: ₹12,500 – ₹37,500 INR
I’m driving an organisation-wide push to personalise every digital touch-point and I need an AI/ML lead who can architect, prototype, and iterate a truly hybrid recommendation engine. Your core domain is Recommendation Technology—specifically the fusion of content-based and collaborative filtering—but you should also be comfortable pulling in signals from NLP pipelines, computer-vision models, or LLM embeddings whenever they strengthen relevance.
Here’s the landscape you’ll step into:
• We have product, behavioural, and metadata streams ready for feature engineering.
• A scalable cloud stack (Python, TensorFlow/PyTorch, Spark, Kubernetes) is in place, waiting for a robust model layer.
• Cross-functional teams are prepared to A/B test whatever you deploy.
What I need from you:
1. A technical roadmap outlining algorithms, data needs, and evaluation metrics.
2. An initial proof-of-concept hybrid recommender that outperforms our current baseline across precision, recall, and diversity.
3. A production-grade implementation with automated retraining, monitoring, and explainability hooks.
Acceptance criteria for phase one:
• Offline validation shows ≥10 % lift in MAP over baseline.
• Real-time inference latency under 150 ms at 99 th percentile.
• Clear documentation of model architecture and feature pipeline.
If you thrive on end-to-end ownership and enjoy collaborating closely with product and engineering, let’s talk and start shaping an experience our users will love.
Here’s the landscape you’ll step into:
• We have product, behavioural, and metadata streams ready for feature engineering.
• A scalable cloud stack (Python, TensorFlow/PyTorch, Spark, Kubernetes) is in place, waiting for a robust model layer.
• Cross-functional teams are prepared to A/B test whatever you deploy.
What I need from you:
1. A technical roadmap outlining algorithms, data needs, and evaluation metrics.
2. An initial proof-of-concept hybrid recommender that outperforms our current baseline across precision, recall, and diversity.
3. A production-grade implementation with automated retraining, monitoring, and explainability hooks.
Acceptance criteria for phase one:
• Offline validation shows ≥10 % lift in MAP over baseline.
• Real-time inference latency under 150 ms at 99 th percentile.
• Clear documentation of model architecture and feature pipeline.
If you thrive on end-to-end ownership and enjoy collaborating closely with product and engineering, let’s talk and start shaping an experience our users will love.