Devyani Tyagi

At Programming Pathshala, I took full ownership of designing and shipping the backend for a DSA Recommender system. I processed 2,500+ unique problems into 384-dimensional semantic vectors for personalized recommendations.

I built the recommendation engine in Python and FastAPI using cosine similarity, and independently architected a decoupled Node.js and MongoDB data service. The system used microservices architecture and was designed for fault tolerance and horizontal scalability.

On my Build With Decency project, I owned the security architecture, combining Google OAuth 2.0 and JWT authentication with Redis-based OTP caching and token blacklisting. I also implemented rate limiting and injection prevention; the system maintained sub-200ms latency while visualizing 4,000+ live data points.

I also built a Python and scikit-learn pipeline for hallucination detection in LLM outputs, achieving 87% accuracy on a 500-sample benchmark. I’m seeking SDE-1 roles where I can build scalable software and grow in a high-ownership engineering culture.