At KundliAI, I shipped critical backend fixes for a WhatsApp AI platform using TypeScript, Supabase Edge Functions, Vertex AI, QStash, Redis, and Razorpay.
I redesigned the donation payment flow to generate unique per-user Razorpay links, fixing credit misattribution. I also reduced webhook latency from roughly 600ms to 400ms by parallelizing four resource-bound pipeline steps.
My security research includes a conference-published IoT intrusion detection system trained on 550K+ BoT-IoT samples, reaching 99.67% accuracy with 14ms inference. I validated it on Edge-IIoT and added SHAP and LIME explanations for per-alert forensic analysis.
I've also deployed a live Telegram behavioral honeypot for scam detection, using Python and HistGradientBoosting to flag scams early. I'm Google Cybersecurity Professional and CCNA certified.

