I'm Naga Durga Rao β a freshly graduated AI & CS engineer from Hyderabad who builds things that actually work, not just things that look good in a notebook.
My work lives at the intersection of Machine Learning and backend engineering. I've trained custom PyTorch models, fine-tuned LLMs with LoRA, built multi-agent pipelines, and wired all of it into production-grade REST APIs β end to end, solo.
I don't wait to "feel ready" before building. Every project in my portfolio started as a real problem I wanted to solve and ended as a system I could defend technically.
π Β B.Tech in CS & AI β Pydah College of Engineering, 2026
π Β Based in Hyderabad, India
π Β Actively looking for AI/ML Engineer or Backend Developer roles
π§ Β Currently deepening: Spring Boot Β· Advanced Django Β· LLM Agents
β‘ Β I learn faster by shipping than by studying
Languages
AI & Machine Learning
Backend & APIs
Frontend
Databases
Tools & Platforms
End-to-end AI healthcare platform with concurrent multi-agent inference
- Built a concurrent multi-agent pipeline running 4 specialized AI agents in parallel β ~60% reduction in inf 8000 erence latency
- Custom PyTorch disease classifier integrated with a scalable Django REST API backend
- OpenCV + DocTR OCR for processing medical documents and extracting structured data
- Firebase for authentication and cloud storage across the full stack
Stack: Python Django PyTorch OpenCV DocTR Firebase ThreadPoolExecutor REST API
Semantic search + multi-LLM backend for competitive exam preparation
- FAISS vector search over UPSC study materials enabling semantic retrieval at scale
- Modular FastAPI backend supporting multiple LLM providers: Groq (Llama 3.1), Gemini, local GGUF models
- Automated news classification using semantic embeddings with no manual labeling
- Open-source β production architecture, not a notebook demo
Stack: Python FastAPI FAISS LangChain Groq API Gemini API Hugging Face Vector DB
LoRA fine-tuned LLaMA model specialized in Data Structures & Algorithms explanations
- LoRA fine-tuning on LLaMA-3B using PEFT for parameter-efficient training
- Systematic prompt engineering + iterative evaluation cycle to improve response accuracy
- Modular ML pipeline following production-grade reproducibility standards
- Deployed to Hugging Face model hub
Stack: Python PyTorch Hugging Face Transformers LoRA PEFT LLaMA-3B
| Certification | Issuer |
|---|---|
| π‘ Salesforce AI Associate | Salesforce |
| π΄ Oracle Generative AI Professional | Oracle |
| π΄ Oracle AI Vector Search Professional | Oracle |
| π΅ Python for Data Science | IBM / EdX |
| π’ Machine Learning Specialization | Andrew Ng / Coursera |
| π’ Generative AI for Everyone | Andrew Ng / DeepLearning.AI |
| π΅ Deep Learning & Neural Networks with Keras | IBM / Coursera |
π Spring Boot β Building production-grade Java microservices
π Advanced Django β REST APIs, Celery, async task queues
π SQL Optimization β Query performance, indexing, schema design
π Generative AI β LLM agents, RAG architectures, fine-tuning pipelines
- π€ Open-source AI/ML projects β LLMs, RAG, agents
- π οΈ Full stack web development collaborations
- π‘ Startup ideas and early-stage technical co-building
- πΌ Freelance AI/ML or backend engineering work
- π Research projects in applied ML
"The best way to learn is to build something real."
