LLM & GenAI Engineer | M.S. in Data Science & Analytics (SUNY Polytechnic Institute)
Graduate Research Assistant @ AIX Lab — Building Production RAG Pipelines, Multi-Agent Orchestration Systems, and Statistical LLM Evaluation Frameworks.
LLM and GenAI Engineer with 3+ years of combined industry and research experience designing retrieval architectures, running systematic LLM experiments (Design of Experiments on GCP H100 VMs), and shipping end-to-end agentic applications:
- Synthetic RAG Benchmarking (
RAG Spectrum): Engineered a 25-category (D01–D25) synthetic RAG challenge generator with multi-layer reliability and exact/fuzzy/semantic deduplication (React + FastAPI). Shipped on Hugging Face Spaces and GitHub. - Statistical RAG Evaluation (
Manthan): Built a full-stack RAG evaluation and experiment-optimization platform using Design of Experiments (DoE) — fractional factorial designs, blocked ANOVA, Tukey HSD, and bootstrap confidence intervals. - Governed Enterprise RAG (
RAPID): Architected privacy-first multi-agent RAG platforms with pre-LLM data firewalls, role/row/column-level access control (RBAC), hybrid Qdrant/FAISS + BM25 retrieval, and audit logging. - AI Workforce Governance (
Pramana): Designing deterministic execution gates, policy enforcement, and verifiable lineage for autonomous agent workflows.
- Histopathology Osteosarcoma Image Classification
Lecture Notes in Networks and Systems (LNNS), Springer Nature Singapore, 2023
Ayush Chhoker, Kunlika Saxena, Vipin Rai, Vishwadeepak Singh Baghela
[Paper Link] · [Google Scholar Profile] · [ORCID]
| Category | Core Technologies & Methods |
|---|---|
| LLM Engineering | RAG Architectures, Multi-Agent Orchestration, LangChain, LlamaIndex, LangGraph, LLM Evaluation, Embeddings, Hybrid Retrieval (Qdrant, FAISS, BM25), Cross-Encoder Reranking, Prompt Injection Defense |
| ML & Statistics | Design of Experiments (Fractional Factorial, Blocked ANOVA, Tukey HSD, Bootstrap CI), Scikit-learn, XGBoost, TensorFlow, PyTorch, PyTorch Geometric (GNNs), SHAP, Statsmodels |
| Languages & Data | Python, SQL, TypeScript, React, Pandas, NumPy, Feature Engineering, Data Cleaning, ETL Pipelines |
| MLOps & Cloud | FastAPI, Docker, Docker Compose, Nginx, GCP (H100 Multi-Model Inference), Ollama, vLLM, Streamlit, Hugging Face Spaces, API Monitoring, GitHub Actions CI/CD, Git |
| Databases & Vector Stores | Qdrant, FAISS, ChromaDB, PostgreSQL (Azure), SQLite, Azure Cosmos DB |
| Project | Domain | Architecture & Key Outcomes |
|---|---|---|
| RAPID | Privacy-First Governed RAG | Enterprise multi-agent RAG with pre-LLM data firewall (ALLOW / ANONYMIZE / BLOCK), hybrid Qdrant/FAISS + BM25 retrieval, read-only SQL agent, confidence-gated routing, and React + FastAPI + Docker + Nginx stack. |
| Manthan | RAG Evaluation & DoE Platform | Full-stack RAG evaluation & experiment optimization applying Fractional Factorial DoE, blocked ANOVA, Tukey HSD, and bootstrap CI to benchmark retrieval/generation pipelines and cost. |
| RAG Spectrum | Synthetic Benchmark Engine | Generates 25 challenge dataset types (D01–D25) with exact/fuzzy/semantic deduplication, evaluation suite, and async batch processing for OpenRouter & Ollama (React + FastAPI) · [Live Demo]. |
| Pramana | AI Workforce Governance | Deterministic trust and execution foundation for autonomous agents with policy gates and verifiable execution lineage. |
| ARIA | Agentic Research Intelligence | Multi-agent research workspace with multi-query expansion, PDF ingestion, source credibility scoring, and traceable evidence chains. |
| CardioFusion | Clinical ML & Explainable AI | Hybrid ensemble cardiovascular risk platform (XGBoost, Random Forest, Neural Net) with SMOTE balancing (95.2% F1, 98.6% ROC-AUC) and SHAP explainability (Live App). |
| Research Compass GNN | Graph Neural Networks | Paper topic classifier & link predictor comparing GAT, GCN, and GraphSAGE on OGB arXiv (169K papers) and AMiner citation graphs (Live App). |