Final-year B.Tech CS student building in AI/ML and computer vision. I like shipping things that work end to end, not collecting half-finished repos.
š Open to SWE / AI Engineer roles š Portfolio: ananaya.vercel.app š¼ LinkedIn: www.linkedin.com/in/ananaya-cse Ā· š:
vision-incident-agent š§ (in progress) Edge-to-cloud CV security system. YOLOv11 + ByteTrack track people on-device, a polygon geofence triggers a 5-second ring buffer, and only 3 keyframes go to Gemini Flash for a structured incident report (threat level, breach type, root cause, recommendations). Sends keyframes instead of streaming video, so token cost drops by over 99%.
Pneumonia detection from chest X-rays Custom CNN with Grad-CAM, a CBAM attention hybrid, and ResNet50 transfer learning, compared side by side. Built during my ML internship at BIT Sindri.
OpenCV 5 DNN field test Benchmarked the new OpenCV 5 DNN engine against the classic one on ONNX models: 5 rounds, 300 measurements, with latency and output-parity analysis.
SpeakLab AI-assisted platform for practicing engineering communication, with interview prompts, prep timers, and audio/video recording. React + Vite + Tailwind, deployed on Vercel.
ML/CV: Python Ā· YOLOv11 Ā· OpenCV Ā· ONNX Ā· MobileNetV2 Ā· ResNet50 Ā· MONAI Web: React Ā· Vite Ā· Tailwind Ā· Three.js Mobile: Flutter Ā· Kotlin
- Software dev intern, CSIR-CIMFR (resource booking system)
- ML/deep learning intern, BIT Sindri
- DoraDao GWY Fellow