I BUILD AGENTIC PIPELINES,
PREVENT HALLUCINATIONS,
GENERATE INSIGHTS.
I am an AI Analyst and Agentic AI Engineer finishing a Master of Science in Information Systems at San Diego State. I sit where business analysis meets applied AI: translating stakeholder questions into multi-agent pipelines, governed analytics, and dashboards leadership actually trusts.
Across research and industry I have shipped LangGraph multi-agent systems, retrieval pipelines with citation guardrails, and BI solutions that cut analyst work by 40 to 60 percent. I care about evals, refusals, observability, and knowing when a model should not answer.
Full-stack AI operations assistant (FastAPI + Next.js) with a natural-language interface over business data. Integrates Claude via tool-use for SQL generation, anomaly flagging, and KPI narration. Deployed on Railway + Vercel with PostgreSQL.
Agentic BI layer that converts raw metric tables into executive-ready narratives. Five-agent LangGraph pipeline: ingestion, anomaly detection, trend analysis, benchmark comparison, GPT-4o narration. Cuts analyst report time by ~60%.
Five-agent executive talent research pipeline built as an interview demo for SPMB Executive Search. Agents: Query Planner, Web Searcher, Profile Extractor, Scorer, and Report Generator producing client-ready HTML briefs. Built with LangGraph + Claude.
Production RAG system with LLM-as-judge evaluation and governance enforced at the architecture level. Features hybrid retrieval (dense + sparse), Ragas-scored faithfulness (0.91) and answer relevancy (0.88), and a HITL review gate before responses are surfaced.
Azure-native multi-agent RAG for semiconductor technical documentation intelligence. Built in 72 hours for an ASML-targeting consulting role. LangGraph + Azure AI Search + Azure OpenAI, LLM-as-judge evaluation, HITL gates, 65 passing tests.
Financial analytics layer that ingests ledger-format CSVs and produces governed insight cards: variance analysis, anomaly flagging, and natural-language P&L summaries. Governed output architecture — every claim is traceable to a source row.
San Diego State University · San Diego, CA
DY Patil College of Engineering · Navi Mumbai, India
LLM2Manim, Arxiv 2026 · FULE Methodology, Multidisciplinary Review 2025
Published research on converting natural-language STEM questions into Manim animations via a 13-node LangGraph pipeline with Docker sandbox execution and governance controls. ManimatedAI capstone, SDSU.
Read on arXiv ↗Systematic review examining a hybrid approach combining FULE methodology and usability heuristics to optimize website usability, reflecting user behavior and perception patterns.
Read on Multidisciplinary Review ↗Open to AI Analyst, AI Engineer, and forward-deployed roles, plus interesting freelance and collaboration. I reply fast.