engineer = {
"name" : "Buvananand Vendotha",
"base" : "Hyderabad, India 🇮🇳",
"focus" : ["Backend Engineering", "Agentic AI", "AI Systems"],
"exploring": "Agentic AI & AI-driven workflows alongside backend engineering",
"seeking" : "Full-time · Collabs · Open Source",
}I design systems that think from RAG pipelines and agentic AI workflows to production-grade backend services. My work lives at the edge of scalability and intelligence.
|
Stateful RAG Narrative Engine · → Repo FastAPI backend orchestrating five AI systems at once LLM narrative generation, a custom 3-layer RAG memory engine (Narrative Relevance Score), local speech-to-text, voice-cloned TTS, and AI image generation.
|
AI Food Delivery Platform · → Repo Full-stack Django platform with AI meal personalization, multi-role dashboards (Customer / Restaurant / Delivery), real-time order tracking, and secure payments.
|
|
Movie Review & Reservation System · → Repo Enterprise-grade Spring Boot REST API combining movie reviews with theater seat reservations JWT auth, RBAC, Stripe payments, PDF receipts, and email notifications.
|
Deepfake Detection Engine · → Repo Spatial frequency deepfake detection combining RGB features with DCT frequency-domain signals through a dual-stream CNN, served via a Flask API.
|
|
Referral & Credit System · → Repo Full-stack referral and rewards platform unique referral links, dual credit rewards on first purchase, fraud prevention, and a real-time metrics dashboard.
|
Autonomous Blog Writer · → Repo LangChain ReAct agent that self-directs research via Wikipedia + web search tools, then writes context-aware content with zero manual prompting. Fully autonomous end-to-end.
|
|
Consequence-Weighted Retrieval for Interactive Narrative · → Repo Research paper introducing NRS, a retrieval scoring function that augments cosine similarity with consequence weighting and recency decay to keep story-critical events retrievable in RAG-based interactive fiction. Implemented in Kalapatha. |
Dual-Stream Spatial-Frequency Deepfake Detection · → Repo Research paper proposing a dual-stream network that fuses RGB spatial features with DCT frequency-domain signals to catch GAN artifacts that spatial-only detectors miss. Implemented in DeepDetect V2. |
AI & Machine Learning
Backend & APIs
Databases & Infrastructure
Frontend
| 🤖 | Exploring agentic AI and AI-driven workflows alongside backend engineering |
| 🌱 | Exploring MLOps, Kubernetes, and distributed system design patterns |
| 🤝 | Open to full-time roles, research collabs & open source contributions |
"The best systems feel invisible until they don't exist."
Let's build something worth remembering.
