AI/ML Engineer · LLM Systems · RAG & Agentic AI · M.Eng UCI '26
LinkedIn · Google Scholar · bghali@uci.edu
| Metric | Result |
|---|---|
| 🧪 LLM eval graders — agreement w/ human evaluators | 94.7% |
| 📈 Learning-to-Rank mAP improvement (175K+ events) | 62% → 94% |
| ⚡ RAG latency reduction (2,500+ daily queries) | −18% |
| 💸 LLM inference cost reduction | −22% |
| 📄 IEEE publications | 6 · 9+ citations |
🔬 AI_Evals Active · UCI M.Eng Capstone × IntuigenceAI
LLM evaluation framework stress-testing 4+ domain-specific agents across document parsing, time-series reasoning, tone/conversation, and end-to-end task completion.
- Built 35+ high-signal graders combining rule-based validators and LLM-as-a-judge workflows
- Achieved 94.7% agreement with human evaluators
- Implemented failure-mode taxonomy (reasoning errors, retrieval failures, tool misuse) with severity scoring
Python DSPy LangChain OpenAI LLM-as-a-judge
🤖 Ad_Placement_Optimization_using_RAG_and_LLMs Multi-agent · RAG
Multi-agent system for targeted ad campaign generation using retrieval-augmented generation.
- 80% Top-5 relevance on campaign targeting
- FAISS vector search + LangChain orchestration + DALL-E multimodal output
Python LangChain FAISS DALL-E OpenAI Cohere
✍️ Text-Summarizer-DSPy Prompt Optimization
Programmatic prompt optimization pipeline using DSPy — comparing automated vs. manual prompt engineering for quality and inference cost.
Python DSPy OpenAI API
📁 CrescentHarborDirectFiler Agentic Pipeline
Automated document filing agent with intelligent routing, classification, and structured output validation. Production-ready with error handling and fallback logic.
Python LangGraph LLM agents
GenAI & LLMs
RAG LangChain LangGraph DSPy FAISS Pinecone MCP LoRA fine-tuning multi-agent systems FSM guardrails Hugging Face Transformers
ML & Data
PyTorch XGBoost LightGBM PySpark feature engineering Learning-to-Rank model evaluation
Cloud & Infra
Azure ML Databricks Docker CI/CD GitHub Actions
Languages
Python SQL Java JavaScript Bash
AI Engineer Intern · Hirello.ai Jan 2026 – Mar 2026
Generative AI career-coaching platform · Gemini (text + voice) · Pinecone · MCP
- Designed FSM-based guardrail architecture — cut agent error rate 10% across 300+ live users
- Orchestrated multi-agent pipelines across 8+ MCP endpoints with RAG-based context memory (Pinecone)
- Optimized multimodal prompt pipelines for real-time Gemini voice + text interactions
ML Research Intern · Colorado State University Jun 2025 – Jul 2025
Behavioral ML systems · Prof. Ortega · NUI Labs · AR notification prioritization
- Boosted Learning-to-Rank mAP from 62% → 94% on 175K+ behavioral events (XGBoost, LightGBM)
- Built scalable ELT pipelines on Azure ML — batch + real-time ingestion, 100% app coverage
GenAI Intern · Indegene Jul 2024
Life sciences AI platform · 2,500+ daily LLM queries
- −18% RAG latency via FAISS vector search + semantic chunking
- −22% LLM inference cost using DSPy-based prompt pipelines
- −30% fine-tuning iteration time via LoRA parameter-efficient fine-tuning
🎓 Master of Engineering in Computer Engineering (Machine Learning and Data Science) · UC Irvine Sep 2025 – Jun 2026
🎓 Bachelors of Technology in Computer Science & Engineering (Artificial Intelligence) · Amrita Vishwa Vidyapeetham 2021 – 2025
Graduate commencement speaker · Final year: Study Abroad at UC Irvine
- 📄 Comparative Analysis of Bitcoin Price Prediction Models: LSTM, BiLSTM, ARIMA and Transformers — IEEE 2024 · 9 citations
- 📄 Scalable Web Crawling: Harnessing Hadoop MapReduce in a Distributed Framework — IEEE 2024
Microsoft Azure AI Engineer (AI-102) Azure Fundamentals (AZ-900) GitHub Foundations Oracle: AI Vector Search