Building scalable backend systems and training LLMs from first principles
Focus Areas
- Backend Engineering (Java, Spring Boot, Microservices, REST)
- Distributed Systems & Performance Optimization
- Large Language Models (LLMs) & Transformer Architectures
- Retrieval-Augmented Generation (RAG)
- ML / AI Systems Engineering with PyTorch
🔗 https://github.com/supraja777/GPT-From-Scratch
Implemented a GPT-style language model from scratch using PyTorch, covering:
- Byte Pair Encoding (BPE) tokenization
- Multi-head self-attention & transformer blocks
- Positional embeddings
- Autoregressive training & inference pipelines
Built to deeply understand transformer internals rather than relying on abstractions.
🔗 https://github.com/supraja777/All-RAG-Techniques
Engineered scalable Retrieval-Augmented Generation pipelines, including:
- Dense embeddings & vector search
- Semantic chunking strategies
- Query rewriting & reranking
- Contextual compression
Focused on improving retrieval quality and reasoning accuracy in LLM systems.
🔗 https://github.com/supraja777/Dual-Agent-Debate-Pattern
Designed a multi-agent LLM debate framework featuring:
- Agent orchestration & state-driven workflows
- Memory management
- Prompt engineering
- Multi-round reasoning and synthesis
Explores structured reasoning and controlled argument generation using agents.
