RAG Retrieval Assistant
Ask any question about Amal's career or select a pre-indexed query below to test the vector matching system.
Professional Summary
Applied AI / Data Engineer with 3+ years of experience at the intersection of data engineering, AI evaluation, and LLM-powered agentic systems. Strong background in building Python-driven evaluation pipelines on GCP (Vertex AI, BigQuery), managing large-scale datasets, and benchmarking LLM and agent architectures for enterprise use cases. Currently an AI Forward Deployed Engineer at Nestingale, deployed to IQ-EQ, building FastAPI and Docker backend services that integrate Azure LLM APIs for loan administration, invoice processing, and document extraction, along with RAG, LangGraph, and MCP-based tooling.
Work Experience
AI Forward Deployed Engineer
Nestingale (Contractor) – Deployed at IQ-EQ
- Build backend services integrating Azure-hosted LLM APIs into a loan administration application.
- Develop an extraction engine and invoice processing workflows that turn unstructured financial documents into structured data.
- Design and containerize RESTful APIs with Python and FastAPI, deployed with Docker.
- Build RAG pipelines and LangGraph agent workflows, with MCP servers exposing tools to the LLM for grounded, document-based answers.
Associate Software Engineer (AI Evaluation & Benchmarking)
Emergence.ai
- Built and maintained Python-based evaluation pipelines for LLMs and agentic AI systems on GCP Vertex AI.
- Evaluated enterprise-grade agent architectures including GenAgents, Web Agents, and Text-to-SQL agents, focusing on correctness, reasoning depth, and robustness.
- Implemented LLM-as-a-Judge frameworks for scalable qualitative evaluation and regression testing.
- Owned benchmarking datasets in BigQuery, designing schemas for evaluation inputs and outputs and enabling structured cross-model and cross-prompt comparisons.
- Analyzed large-scale evaluation data to identify performance bottlenecks, failure modes, and optimization opportunities.
- Partnered with research and engineering teams to align evaluation pipelines with deployment and production needs.
- Improved benchmarking scalability, reproducibility, and reliability across multiple agent types and workflows.
Annotation Analyst
Merlyn Mind
- Conducted inter-annotator agreement (IAA) analysis using Python to measure consistency and identify ambiguous cases.
- Implemented majority-vote and consensus logic in Python to resolve annotation conflicts and finalize ground-truth labels.
Featured Technical Projects
Citadel-Archival-Search
Python • LangGraph • FastAPI • Qdrant • Neo4j • Self-RAG
An enterprise-grade Self-Corrective RAG (Self-RAG) system for querying A Song of Ice and Fire lore using a cyclic LangGraph state machine. Features hybrid search, concurrent document relevance grading, hallucination detection, and query optimization loops.
Scripbox-KB-Retriever
Python • RAG Architecture
A robust knowledge base retrieval system built utilizing advanced RAG architectures and Python for parsing large-scale contexts.
pdf-qa-retriever
Python • Vector Embeddings
A document parsing and question-answering tool leveraging vector embeddings for accurate information retrieval.
Research & Publications
Emergence WebVoyager: Toward Consistent and Transparent Evaluation of (Web) Agents in The Wild
Deepak Akkil, Mowafak Allaham, Amal Raj, et al. • arXiv, 2026 • arXiv:2603.29020
Provided high-quality human-in-the-loop annotations, logic verification, and rigorous data quality validation for multimodal agentic workflows, contributing to the foundational evaluation benchmark datasets.
Read paper on arXivLLM & Agentic Evaluation Benchmarks
arXiv:2410.00689 • arXiv:2407.13032
Provided high-quality human-in-the-loop annotations and logic verification for multimodal agentic workflows, contributing to validation datasets of published research.
Education
BTech, Chemical Engineering
OPJS University, Churu, RJ