name: Ayush Kumar
alias: ayushsyntax
role: Data Scientist & Machine Learning Engineer
location: India
education: B.Tech — Artificial Intelligence & Data Science
university: Kurukshetra University
what_i_build:
- Agentic AI systems with persistent memory & observability
- End-to-end RAG pipelines with dynamic query reasoning
- Reproducible MLOps: MLflow · DVC · Docker · AWS
- Transformers & deep learning architectures from scratch
currently_working_on: Reasona — a self-reflective RAG engine
motto: "Code like a wizard, deploy like a boss."
|
LangGraph agent with persistent memory, modular tool integration, and LangSmith tracing. Human-in-the-loop for controlled autonomous research. |
Time-series forecasting with MLflow, DVC, Docker, and AWS EC2/S3. Fully reproducible, production-ready MLOps stack. |
|
Decoder-only Transformer trained on Shakespeare, reaching 3.02 perplexity. Custom attention, tokenization, and training loop — no shortcuts. |
Self-reflective RAG engine inspired by HyDE and SEAL. Dynamic query expansion and multi-stage reasoning for grounded, accurate retrieval. |
Languages
Data Science
ML / Deep Learning
GenAI & LLMs
MLOps & Deployment
