Building
Intelligent
Systems
with
Code
and
Curiosity
AI Engineer with 3+ years across industry engineering, zero-to-one product development, graduate AI research, and applied data science. I build agentic systems, retrieval workflows, AI tools, and healthcare ML applications with a focus on reliability, evaluation, and measurable product impact.
Artificial Intelligence Through an Engineering Lens
My work sits at the intersection of machine learning, data systems, and scalable engineering. My graduate studies in Computer Engineering at George Washington University focused on Machine Learning and Intelligent Systems, where I worked on building practical ML systems that move from research ideas to deployable software.
I approach machine learning as an engineering discipline and am interested in the full lifecycle of intelligent systems — from data pipelines and experimentation to infrastructure, deployment, and iteration in production environments.
My work involves designing end-to-end systems that combine classical machine learning, deep learning, and modern AI tooling, including recommendation systems, large language model applications, and cloud-based ML infrastructure. I am particularly interested in the intersection of machine learning systems, generative AI, and scalable data infrastructure.
I enjoy building systems where data, engineering, and machine learning come together to create practical intelligence.
Technical expertise
Selected projects
From deployed products to active research and new initiatives.
MARS: Multi-Agent Research Intelligence System
A deployed research intelligence platform that coordinates seven specialized agents across planning, retrieval, tool execution, evidence synthesis, question answering, citation, and reporting. MARS combines multi-document RAG, metadata-aware Qdrant retrieval, structured outputs, authentication and RBAC, citation validation, hallucination guardrails, saved runs, monitoring, and offline evaluation in a production-style application.
Explore MARS
MARS was created to reduce the time required to read, compare, and cite technical literature. Its workflow uses seven specialized agents to plan research tasks, retrieve evidence, call tools, synthesize findings, validate citations, check unsupported claims, and generate structured reports.
The platform includes document ingestion and chunking, embeddings, metadata filtering, Qdrant vector search, persistent workflow state, external LLM integrations, authentication, error handling, usage and cost visibility, logging, monitoring, and offline evaluation. The current product is deployed and available for live demonstration.
Thoracic Disease Detection System
An end-to-end multi-label medical imaging system for detecting nine thoracic diseases across more than 112,000 NIH chest X-rays. The project benchmarks DenseNet-121 and Vision Transformer models, evaluates disease-level AUROC, analyzes class imbalance and failure cases, and uses Grad-CAM to make predictions more interpretable.
Explore Medical AI Project
The system covers preprocessing, augmentation, multi-label training, validation, inference, disease-level AUROC comparison, class-imbalance analysis, and error investigation. DenseNet-121 and Vision Transformer architectures were compared to understand performance tradeoffs, while Grad-CAM visualizations were used to inspect clinically relevant image regions and make model behavior easier to evaluate.
EREN: Code-Aware Chunking and Retrieval Strategies for High-Level Synthesis Code Optimization
A retrieval-augmented generation research project for High-Level Synthesis code optimization and quality-of-results improvement. EREN studies how code-aware chunking, HLS-specific indexing, hybrid retrieval, query translation, HyDE, and reranking affect the quality of generated HLS optimizations across a self-curated corpus spanning 12 benchmark suites.
My contribution: Converting more than 500 heterogeneous C, C++, and TCL sources into clean, validated, retrieval-ready data; standardizing metadata; and building reproducible workflows that supported retrieval and generated-output evaluation.
Explore EREN Research
EREN investigates retrieval strategies for producing better High-Level Synthesis code optimizations. The research corpus spans 12 HLS suites, including CHStone, DP-HLS, DSS, Inter-Block, MachSuite, PNA-Analyser, Vitis_Accel, flowgnn, forgebench, gnnbuilder, hyle, and polybench.
The study compares token-based and code-aware chunking, generic and HLS-specific indexing, LLM-only and vector baselines, hybrid and mixed retrieval, query translation, HyDE, and optional reranking. Evaluation uses Vitis HLS 2024.2 and examines initiation interval, latency, and resource-utilization measures. The research found that code-aware retrieval strategies produced stronger quality-of-results outcomes than naive vector RAG and prompt-only baselines.
Low-Latency Voice AI Agents for Recruiting
An in-development real-time Voice AI Agent system for recruiting workflows, designed around the lowest practical conversational latency. The planned system combines streaming speech recognition, fast model inference, streaming text-to-speech, voice activity detection, turn-taking, interruption handling, and latency observability to create responsive, natural conversations.
Explore Voice AI Architecture & Roadmap
The goal is to build a production-oriented voice agent that can support recruiting conversations without the long pauses common in basic voice bots. The architecture will optimize each stage of the real-time loop: audio streaming, endpoint detection, transcription, context assembly, model response, speech synthesis, playback, and interruption recovery.
The project will measure time to first transcript, model time to first token, time to first audio, end-to-end response latency, interruption recovery, and conversation completion quality. It will also explore telephony integration, session state, failure handling, logging, and human handoff patterns.
Experience
Data Science Fellow
- Contribute to Data, AI, and Agentic AI initiatives for nonprofit and community-centered use cases, translating stakeholder objectives into scalable data workflows, applied analytics, ML solutions, and intelligent applications.
- Develop reproducible Python implementations, evaluation workflows, and technical documentation while assessing data quality, model behavior, reliability, and practical delivery constraints.
Graduate AI Researcher — RAG Systems & Evaluation
- Architected a retrieval and data pipeline that converted more than 500 heterogeneous C, C++, and TCL sources into clean, validated datasets for an LLM-grounded High-Level Synthesis research system.
- Designed reproducible evaluation workflows to measure retrieval relevance and generated-output quality, trace failure modes, compare prompt and model iterations, and communicate accuracy, latency, reliability, and cost tradeoffs.
Chief Technology Officer & Founding Engineer
- Led zero-to-one development of an AI-enabled recruiting platform, translating recruiter workflows into Python services, REST APIs, document pipelines, data models, automation, and web applications.
- Scaled pipelines to 1,000+ candidate documents per batch, reducing manual shortlisting time by approximately 70%.
- Owned SQL, MongoDB, Redis, search, authentication and RBAC, integrations, deployment, testing, logging, monitoring, production troubleshooting, scalability, and latency improvements.
Principal Consultant — HCM
- Supported workforce management and payroll systems for more than 11 multinational clients, troubleshooting business-critical attendance, leave, payroll, exit, and employee-lifecycle workflows.
- Developed and debugged more than 30 JavaScript and SQL business rules, reports, and automations, improving data integrity, auditability, and operational efficiency.
Academic background
Master of Science, Computer Engineering
Focus: Machine Learning and Intelligent Systems. Coursework: Advanced Machine Learning, Pattern Recognition & ML, Machine Intelligence, Reinforcement Learning, Big Data & Cloud Computing, Network Security.
Bachelor of Technology, Automobile Engineering
Built a strong engineering foundation in systems thinking, control systems, applied mathematics, and structured problem solving before transitioning into machine learning and AI.
Let's build reliable AI systems together.
I am actively exploring AI Engineer, Machine Learning Engineer, Applied AI, and Agentic AI opportunities across the United States. Based in Houston, open to relocation.