I build production-grade, hallucination-resistant AI systems. I specialize in transforming messy, unstructured data into reliable, multi-step agentic workflows and advanced Retrieval-Augmented Generation (RAG) pipelines.
Currently focused on:
- 🤖 Agentic Orchestration: Building multi-tool agents with dynamic routing, self-correction loops, and human-in-the-loop (HITL) breakpoints.
- 🔍 Advanced Retrieval: Implementing hybrid search (Dense + Sparse/BM25) and graph-based retrieval for complex, real-world datasets.
- 📏 LLM Evaluation: Instrumenting observability (Langfuse) and automated evaluation frameworks (RAGAS/DeepEval) to measure and guarantee AI reliability.
| Category | Technologies |
|---|---|
| Languages | Python (Expert), JavaScript, TypeScript, SQL |
| AI / ML | LangChain, LangGraph, Hugging Face, RAGAS, DeepEval, DSPy |
| Backend | FastAPI, Celery, Redis, REST APIs, Webhooks |
| Databases | PostgreSQL, Qdrant (Vector), ChromaDB, Neo4j (Graph) |
| Frontend | React, Next.js, Tailwind CSS, Streamlit |
| DevOps / Tools | Docker, Git, GitHub Actions, AWS, Modal |
An intelligent, multi-tool legal document analysis platform.
- Architecture: Agentic routing classifies queries into extraction, logical reasoning, or search paths.
- Retrieval: Hybrid search combining dense vectors (BGE-M3) and sparse keyword retrieval (BM25) via Qdrant.
- Scale: Asynchronous document processing pipeline using Celery + Redis for heavy PDF ingestion.
🗄️ RAG_SQL
A production-grade, self-correcting Text-to-SQL agent.
- Reliability: Features AST-level syntax validation (
sqlglot) to catch errors before execution. - Agentic Loop: Implements a recursive self-correction mechanism that feeds execution errors back to the LLM to autonomously fix the query.
- Context: Graph-based RAG for intelligent schema retrieval and foreign key analysis.
**
I am currently open to Full-Time AI/ML Engineering and Software Engineering (AI) roles where I can build scalable, impactful agentic systems.
- 📧 Email: [email protected]
- 💼 LinkedIn: Agboola Aaliyah
- 🌐 Portfolio/Website:
“The best way to predict the future of AI is to build reliable systems that work today.”