Backend Systems • AI/ML • Legal AI • Research
I build backend systems and AI applications that actually work in real-world conditions — not just demos.
My work currently focuses on:
- Retrieval-Augmented Generation (RAG)
- Domain-specific LLM systems
- Backend architecture for scalable applications
- Research-driven system design
A system-level implementation of a legal AI assistant designed for real-world legal workflows.
-
Built on a RAG pipeline over large-scale legal corpora
-
Focus on retrieval accuracy, not just generation
-
Designed improvements around:
- structured chunking
- category-aware search
- reranking strategies
-
Integrated LLM fine-tuning (LoRA / PEFT) into the pipeline
This project is an ongoing system build — not a static prototype.
An independent research direction exploring a new representation framework for probabilistic systems.
- Based on Fourier-algebraic transformations
- Focus: constraint-embedded inference and system modeling
- Work includes formulation, structure design, and theoretical validation
- Manuscript in preparation
This is long-term research, not a one-off project.
Backend Systems → Node.js, Java, APIs, data flow
AI / ML Systems → RAG, LLM fine-tuning, embeddings
Data & Retrieval → Vector DBs, semantic search, indexing
System Design → architecture, failure analysis, iteration
- Contributor to Ganga Project
- Fixed production-level bug (Python NameError)
- PR merged after debugging and validation
- Improving retrieval quality in AI systems
- Bridging backend engineering with LLM workflows
- Expanding research into usable systems