╔══════════════════════════════════════════════════════════════════╗
║ AI SYSTEMS ENGINEER · RAG · MULTI-AGENT · OSS ║
╚══════════════════════════════════════════════════════════════════╝
Building production-grade LLM infrastructure — pipelines, provenance and agents that work in the real world.
profile = {
"name" : "Balaji Jegadeesh Vijayaraghavan",
"role" : "AI Systems Engineer",
"focus" : ["RAG Pipelines", "Multi-Agent Orchestration", "LLM Evaluation"],
"education" : "MS Computer Science @ Binghamton University (May 2026)",
"currently" : ["SWE @ ProMed Health+", "Research Assistant @ Watson College"],
"published" : "HopeLink · IISE Annual Conference 2026",
"open_to" : "AI Engineering / Agentic AI roles / Full Stack Developer roles · Summer 2026",
}🔷 Dokis · pip install dokis
RAG Provenance Middleware - verifies every LLM claim is grounded in a retrieved source
Query → Retrieval → [DOKIS LAYER] → Grounded ✓ / Ungrounded ✗ → Audit Log
↑
~1ms · No LLM call · JSON audit log
LangChain adapter · LlamaIndex adapter
86-test CI suite · Python 3.10–3.12
🔶 HopeLink · IISE 2026 · Peer-Reviewed
Domain-Restricted Clinical RAG — oncology patient support with verified provenance
User Query
│
▼
┌─────────────────────────────────────────────┐
│ RESEARCH PHASE │
│ Allowlist: NCI · PubMed · NCCN │
│ URL-level enforcement · Citation logging │
└──────────────────┬──────────────────────────┘
│ 85% citation compliance
▼
┌─────────────────────────────────────────────┐
│ SYNTHESIS PHASE │
│ Evidence-grounded generation │
│ Patient-friendly · Conservative safeguards │
└─────────────────────────────────────────────┘
Evaluated across 1,300+ responses · 5-level complexity taxonomy · 6-category failure framework
ReAct Reasoning + Episodic Memory — agents that remember across sessions
Input → [ReAct Loop] → Tool Calling → Retry Logic → Response
↓
Episodic Memory (cross-session)
↓
Automated Evaluation Framework
↓
SSE Streaming UI (Next.js · live reasoning visualization)
LLM SYSTEMS │ RAG · Multi-Agent · ReAct · Tool Calling · LLM Evaluation
│ Episodic Memory · SSE Streaming · Ollama · OpenAI · Anthropic SDK
│
BACKEND │ Python · FastAPI · Django · Flask · Spring Boot · Node.js
│ PostgreSQL · Redis · Async Pipelines · REST APIs · JWT
│
INFRASTRUCTURE │ Docker · Azure · AWS · GCP · CI/CD · GitHub Actions
│
OSS │ PyPI · LangChain · LlamaIndex · BM25 · Provenance Middleware
┌─────────────────────────────────────────────────────────────────┐
│ HopeLink: A Large Language Model–Powered Clinical Information │
│ Assistant Supporting Cancer Patients' Online Information │
│ Seeking │
│ │
│ · IISE Annual Conference 2026 (Accepted · Minor Revisions) │
│ · First Author │
│ · Dual-phase RAG · 85% citation compliance │
│ · 1,300+ evaluated responses · Failure taxonomy │
└─────────────────────────────────────────────────────────────────┘
Nov 2025 – Present Software Developer ProMed Health+
HIPAA-compliant SaaS · 50+ production APIs
Azure · JWT + MFA · QuickBooks integration
Jun 2025 – Present Research Assistant Binghamton University
HopeLink · Clinical RAG · IISE 2026 Publication
Aug 2021 – Sep 2024 Senior Software Engineer LTIMindtree
20,000+ enterprise assets · Data lineage engine
30min → 60sec document validation
╔══════════════════════════════════════════════════════╗
║ Building AI systems that are grounded, auditable, ║
║ and safe — not just fluent. ║
╚══════════════════════════════════════════════════════╝
linkedin.com/in/vbj1808 · pypi.org/project/dokis · bvijayaragha@binghamton.edu