I turn AI into products people actually want.
I’m Vivek Praturi, an applied AI builder working across recommender systems, generative AI, and human-centered machine learning.
Make the complex feel obvious.
AI becomes valuable when the technology disappears into the experience.
Scroll to explore ↓01 / Perspective
My operating principleDon’t start with the model. Start with the moment that should feel better.
My work connects rigorous machine learning with product intuition: understanding the person, the context, and the next useful action.
02 / Selected work
Selected experiments.
Small, opinionated products built to test a sharp idea in the real world.
macOS app · 2026
Attune
A real-time attention feedback layer for children’s learning — on-device webcam sensing that nudges focus back during homework, without recording or uploading video.
- Swift
- Core ML
- On-device
Interview prep · 2026
Work Simulator
Makes interview prep feel alive — simulates a real work environment with teammates, tickets, and tradeoffs instead of grinding LeetCode.
- TypeScript
- AI agents
- Interview prep
Crash course · 2026
AI Crash Course for Product Managers
A fast, practical primer on modern AI for PMs — enough technical depth to ship informed product decisions without becoming an engineer.
- AI product
- Product managers
- Interactive web
Research · 2025
LangMem Benchmark
A benchmarking framework that isolates semantic, episodic, and procedural memory to measure their impact on multi-turn conversational search.
- Python
- Conversational AI
- Benchmarks
03 / Field notes
Thinking in public.
Notes on applied AI, recommendation systems, and what I learn while building.
On-device ML · 8 min
Building Attune’s ML stack (2/2)
Apple Vision, Core ML, and pretrained models for real-time focus — no cloud inference. ↗Attune · 7 min
An attention layer for the Mac (1/2)
Why we’re building gentle focus feedback that sits over whatever your child is already doing. ↗Conversational AI · 6 min
Building conversational AI with memory
Semantic, episodic, and procedural memory in practice. ↗Recommender systems · 1 min
Transformers4Rec: the TL;DR version
A quick path from interaction data to sequential recommendations. ↗More soon
Follow the notebook
Experiments become notes. Notes become better experiments. ↗
Vivek Praturi
Austin, Texas
04 / About Vivek Praturi
Builder by instinct.
Scientist by training.
I work in machine learning and generative AI, with a focus on product recommendations at Nike.
I’m interested in systems that learn from behavior without losing sight of the human behind the data. Outside my day job, I build small, opinionated projects to test ideas quickly and share what holds up.
Recommendations, GenAI, applied ML
Masters in AI · UT Austin
Prototype, measure, explain, repeat
05 / Let’s connect
Have an interesting
problem in mind?
I’m always glad to talk about recommendation systems, applied AI, product ideas, and ambitious experiments.