At Evol Technobits Digital, I integrated Next.js interfaces with backend APIs for client-facing business websites. I also participated in debugging, feature implementation, and code reviews.
I built an AI-Powered Apparel Size Estimation System with a FastAPI WebSockets pipeline that processed 10 FPS video streams with sub-100ms backend latency. It used YOLOv8 and MediaPipe Pose to extract keypoints, with filtering and voting to reduce coordinate jitter.
For InterviewForge, I built a split-generation pipeline that launched interview rooms in under three seconds, and integrated a sandboxed compiler with a local AST-parser fallback. I also trained a Random Forest classifier to predict SDE placement readiness.
On Recall AI, I cut Modal serverless cold-start latency from 20 seconds to 1.2 seconds through lazy-import optimization. I also implemented semantic search over screenshots using Supabase pgvector and Hugging Face embeddings.
