I’m Anay, an AI Engineer focused on GenAI backends and system design.
I don’t just prompt models — I design, evaluate, deploy, and scale AI systems that solve real problems. My work spans retrieval-augmented generation, multimodal pipelines, and high-throughput backend services.
- 🎓 B.Tech CSE @ IIIT Vadodara – ICD (2022–2026)
- ⚡ AI Intern @ Redblox.io
- 🧠 Previously GenAI Backend Intern @ FutureSmart AI
👉 Portfolio: https://anaypandey.vercel.app
- Designing production RAG architectures (not toy demos)
- Multimodal ingestion: PDFs, images, scanned + handwritten docs
- Evaluation loops, hybrid scoring, latency & retrieval tuning
- Async APIs and backend services for AI workloads
- Migrated vector store from ChromaDB → Qdrant, improving stability & search efficiency
- Built backend modules for WorkLog App (auth + API workflows)
- Engineered async RAG pipelines → 40% lower latency
- Improved retrieval precision by 28% across 600+ sessions
- Designed distributed vector indexing & embedding services
- Built iPatronus, a real-time teleconsultation platform (WebSockets + Daily API)
- Designed multimodal document pipeline (Vision LLM + OCR)
- Generated structured clinical summaries using Haystack RAG
Languages
Python · TypeScript · JavaScript · C++ · SQL
GenAI / LLM Systems
RAG · LangChain · Agents · AST Parsing · OCR · Embeddings
Vector Databases
Qdrant · FAISS · Pinecone
Backend & APIs
FastAPI · Node.js · Express · REST · WebSockets · Async Microservices
Frontend (supporting role)
React · Next.js · Tailwind CSS
DevOps
Docker · GitHub Actions · Vercel · Render · Railway
- Repo-level RAG using AST parsing + dependency graphs
- Natural language queries grounded to exact files & line numbers
- Hybrid semantic + structural retrieval
- Agentic system to replicate UI, routing & APIs of target apps
- Automated DOM extraction, component mapping & code generation
- Feedback loops to improve UI + logic fidelity
- Hybrid FAISS embeddings + OCR preprocessing → +42% precision
- Batched async pipelines → 25% faster embeddings
- Supports text, scanned docs & images
- Sub-2s PDF Q&A for 50+ users
- Adaptive chunking & retrieval tuning → +35% accuracy
- Incremental ingestion without full reindexing
- EEG Emotion Recognition (BCICIV_2a dataset, 14+ R visualizations)
- Software-Defined WSN Simulator (HEED, LEACH, GUI, energy models)
- Secure Electronic Voting System (SSL/TLS, 3 CA infrastructure)
- O-RAN DRL-based RAN Slicing (WSL-friendly deployment)
- Distributed & graph-based RAG
- Multimodal evaluation strategies
- DRL for networking & infra systems
- Performance engineering (CUDA, batching, async)
📍 Mumbai, India
📧 anaypandey1504@gmail.com
🌐 https://anaypandey.vercel.app
🐙 https://github.com/anaypandey1504
“I care about systems that work — not demos that just look good.”