AI/ML Engineer specializing in production-grade, full-stack AI systems — Machine Learning model training and evaluation in PyTorch, Agentic RAG pipelines, LLM applications. Final-year B.E. AI & Data Science, VTU (CGPA 7.9/10). Bias toward systems that are secure, well-tested, and run in production, not just a notebook.
RAG · LLMs · LangChain · LangGraph · MCP · Hugging Face · FAISS · XGBoost · SHAP · JWT · Redis · Qdrant · CI/CD · FHIR R4 · OMOP CDM
| Project | What | Links |
|---|---|---|
| ⭐ Aurelia AI — Multi-Tenant AI Customer Support Platform (2026) | Qdrant-backed knowledge bases, SSE streaming chat, RAG over docs + websites, LLM routing across OpenAI/Anthropic/Google/DeepSeek | Demo · Code |
| NexaMind AI — Enterprise Agentic RAG Workspace (2025) | Document chat, streaming AI conversations, multi-step agent workflows, MCP tool calling, Chroma/Milvus | Demo · Code |
| Aurevia Health AI — Clinical Intelligence Platform (2025) | LangGraph multi-agent RAG, CatBoost/XGBoost/TabICLv2 + SHAP, FHIR R4 / OMOP CDM, RBAC + PII redaction | Demo · Code |
| School | Detail | Years |
|---|---|---|
| Visvesvaraya Technological University (VTU) | B.E. Artificial Intelligence & Data Science · CGPA 7.9/10 | 2022 – 2026 |
| Hoysala PU College | Pre-University, Science (PCM) · 78% | 2020 – 2022 |
| SBS School | SSLC · 90.56% | 2019 |
- Hugging Face — LLM Course · Agents Course · Context Engineering · MCP Course (certificates in hand)
- Microsoft Applied Skills — Develop an agent with integrated tools
- DeepLearning.AI & OpenAI — ChatGPT Prompt Engineering for Developers
- Databricks — Generative AI Fundamentals Accreditation
- IBM SkillsBuild — AI Fundamentals · Microsoft Learn — Explore Generative AI
English (Full Professional) · Hindi (Professional) · Kannada (Native) · Telugu (Professional)
⭐ Aurelia AI — what I built & what I used
Multi-tenant AI customer support platform: Qdrant-backed vector knowledge bases, SSE-streamed chat across isolated tenant workspaces, embeddable TypeScript chat widget + admin dashboard, LLM routing across OpenAI / Anthropic / Google / DeepSeek, JWT/bcrypt auth, SSRF-protected ingestion.
Next.js 14 · React 18 · TypeScript · FastAPI · PostgreSQL/SQLite · Redis · Qdrant · Docker · JWT
NexaMind AI — what I built & what I used
Enterprise agentic RAG workspace: real-time streaming AI conversations, multi-step agent workflows across document types and concurrent sessions, parsing → chunking → vector indexing pipeline, MCP tool calling, NextAuth/JWT + RBAC/SSRF hardening.
Next.js 16 · React 19 · TypeScript 5 · Tailwind 4 · FastAPI · SQLAlchemy · Chroma/Milvus · Redis · NextAuth · JWT
Aurevia Health AI — what I built & what I used
Privacy-first clinical decision support: LangGraph multi-agent RAG for real-time clinical workflows, trained CatBoost / XGBoost / TabICLv2 with SHAP explainability and multi-organ digital twin, FHIR R4 + OMOP CDM v5.4 compliance, RBAC + PII redaction.
React 19 · FastAPI · LangGraph · Ollama · CatBoost · XGBoost · SHAP · PostgreSQL/SQLite · Docker · JWT · RBAC
- Portfolio: https://varunbp-portfolio.vercel.app/
- Mail: varunbpvarunbp@gmail.com
- LinkedIn: https://linkedin.com/in/varunbp09
- Location: Nelamangala, Karnataka, India