Resume

I'm an AI research engineer and manager based in Houston, working at the intersection of generative AI, large language and vision-language models, agentic systems, and physics-inspired learning. I currently lead AI strategy and technical research at BKO AI — designing agentic swarm architectures, GraphRAG pipelines, and multimodal systems for industrial applications. Before that I was an AI Research Scientist at Shell and a Ph.D. researcher at NCSU on the James Webb Space Telescope's EIGER Collaboration.

Generative AILLMsVLMsMultimodal AIEmbodied IntelligenceAI AgentsPhysics AIAI for Science
Experience

BKO AI — Houston, TX

Oct 2024 – Present

Strategic Leadership
Senior AI Manager Jan 2026 – Present
AI Technology Manager May 2025 – Jan 2026
  • Executive Strategy: Define and execute the strategic AI roadmap, aligning GenAI, multi-modal, and Knowledge Graph initiatives with business goals to drive innovation and operational efficiency.
  • Product Leadership: Direct the lifecycle of all Agentic AI products on the enterprise platform, including Deep-search, Wide-search, Unified Vision, GUI automation, and Self-executing task agents.
  • Team Management: Manage a cross-functional team of AI researchers, software engineers, and graph specialists; improved project delivery timelines by 30% through Agile implementation and roadmap alignment.
  • MLOps Governance: Established standardized MLOps practices including CI/CD pipelines, model monitoring, and secure Azure-based deployment, ensuring reproducibility and governance.
Technical Core
Principal AI Research Engineer Oct 2024 – Present
  • Agentic Swarm Architecture: Architected an AI agentic swarm system integrated within the Knowledge Graph Explorer, enabling autonomous cross-collaboration between agents for complex query resolution.
  • GraphRAG Implementation: Spearheaded the design of fine-tuned multi-LLM-agents using GraphRAG and Neo4j on Azure. Streamlined CAD blueprint retrieval and budget estimation, improving planning accuracy by 55% and reducing retrieval latency by 40%.
  • Strategic Knowledge Graphs: Designed Context Graphs for Tag-mapping and Failure Mode and Effects Analysis (FMEA), significantly enhancing agent decision-making capabilities and trading signal tracking.
  • Multi-Modal Pipeline: Led the development of a pipeline integrating thermal sensors, IoT devices, and text reports for predictive maintenance. Achieved 92% accuracy, reducing equipment downtime by 40% and preventing $15M in annual losses.

Shell International Exploration and Production — Houston, TX

Oct 2022 – Oct 2024

AI Research Scientist Oct 2022 – Oct 2024
  • Legal AI Automation: Revolutionized contract generation by fine-tuning LLaMA-3.2 with LoRA, RAG, and LlamaParse. Reduced processing time from 2 weeks to 10 minutes, saving $8M annually. Implemented RLHF and an internal reflection mechanism to ensure legal accuracy.
  • 3D Vision at Scale: Developed state-of-the-art multiclass segmentation algorithms using transformer-based LVMs for 10 TB of 4K-resolution CT-scan data. Achieved 0.841 mIoU on severe class imbalances, generating $20M in annual revenue growth.
  • Generative Super-Resolution: Applied diffusion models for rock image super-resolution (up to 32K), accelerating processing from 3 days to 15 minutes (99% efficiency boost).

North Carolina State University — Raleigh, NC

Aug 2019 – Oct 2022

Ph.D. Research Assistant Aug 2019 – Oct 2022
  • JWST Data Analysis: Built deep learning pipelines to analyze multidimensional spectral data from the James Webb Space Telescope (EIGER Collaboration).
  • Bayesian Inference: Reduced measurement errors by 80% in intergalactic medium studies using generative deep learning, achieving reconstruction error rates as low as 0.025%.
  • GUI Development: Created deep learning pipelines and GUI software to analyze multidimensional spectral data from the James Webb Space Telescope, a $10B project exploring the deep Universe, using PyTorch.

Education

Ph.D. in Physics

2022 · North Carolina State University, Raleigh, NC

Concentration: AI in Physics, Generative Modeling

M.S. in Physics

2018 · North Carolina State University, Raleigh, NC

M.S. in Electrical Engineering

2018 · North Carolina State University, Raleigh, NC

Concentration: Computational Intelligence

B.S. in Mechanical Engineering

2015 · North Carolina State University, Raleigh, NC

B.S. in Physics

2015 · North Carolina State University, Raleigh, NC

Skills
Languages & Web
PythonC++C#JavaTypeScriptJavaScriptSQLHTML/CSSMATLAB
Frameworks & Libraries
PyTorchTensorFlowJAXHuggingFaceDiffusersTransformersTRLFastAPIReactViteCUDAScikit-Learn
Agentic AI & Graph
LangChainLangGraphLlamaIndexAgentScopeNeo4jGraphRAG
Cloud & MLOps
AWS (SageMaker, EC2)Azure (AI Studio, DevOps)DockerKubernetesCI/CDGit
Generative AI Tools
ComfyUIGradioGPT-SoVITSOllaman8nDify
Selected projects
  • E-Commerce Generative AI Suite · 2022 – Present — StableDiffusionXL, Flux.1, ControlNet, and SegmentAnything2 composed into a controllable 8K product-image pipeline serving 50+ clients.
  • Virtual Influencer Pipeline · 2022 – Present — GPT-SoVITS + Kling AI system for multilingual virtual influencers, reducing video customization time to 6 hours.
  • Digital Twin with Gaussian Splatting · 2023 – 2024 — Sneaker digital twin reconstruction using NeRF and 3D Gaussian Splatting at 97% geometric accuracy.
  • Autonomous Navigation System · 2020 – 2022 — NVIDIA Jetson, LIDAR, and ZED cameras with vSLAM for 0.05 s end-to-end navigation latency.
Awards & publications
Awards
  • Top 10 Breakthroughs of the Year · 2023 — EIGER Collaboration for the James Webb Space Telescope
Publications
  1. B. Liu, R. Bordoloi. A Deep Learning Approach to Quasar Continuum Prediction. Monthly Notices of the Royal Astronomical Society, Vol. 502, Issue 3, April 2021, pp. 3510–3532.
  2. B. Liu. Ph.D. Thesis: Study of Intergalactic Medium Using Spectroscopy and Photometry with Deep Learning.
  3. R. Bordoloi, B. Liu. rbcodes v0.2: JWST/NIRCam Grism Spectroscopic Analysis API. Zenodo, October 2022.
  4. R. Bordoloi, R. A. Simcoe, J. Matthee, D. Kashino, R. Mackenzie, S. J. Lilly, A.-C. Eilers, B. Liu, D. DePalma, M. Yue, R. P. Naidu. EIGER IV. The Cool 10⁴ K Circumgalactic Environment of High-redshift Galaxies Reveals Remarkably Efficient Intergalactic Medium Enrichment. The American Astronomical Society, Vol. 963, Number 1, p. 28, 2024.
  5. B. Greig, S. E. I. Bosman, F. Davies, D. Durovcikova, H. Fathivavsari, B. Liu, R. A. Meyer, Z. Sun, V. D'Odorico, S. Gallerani, A. Mesinger, Y.-S. Ting. Blind Quasar Reconstruction Challenge: Exploring Methods to Reconstruct the Lyman-Alpha Emission Line of Quasars. Monthly Notices of the Royal Astronomical Society, Vol. 533, Issue 3, September 2024, pp. 3312–3343.
  6. D. Kashino, S. J. Lilly, J. Matthee, R. Mackenzie, A.-C. Eilers, R. Bordoloi, R. A. Simcoe, R. P. Naidu, M. Yue, B. Liu. EIGER VII. The Evolving Relationship between Galaxies and the Intergalactic Medium in the Final Stages of Reionization. The Astrophysical Journal, Vol. 997, Issue 2, id. 280, 26 pp.
Invited talks
  • November 2025 · Houston, TX Grounding AI Agents with Root Cause Analysis beyond MCP Tools — Future AI
  • November 2025 · Houston, TX Failure Mode and Effects Analysis Graph as MCP Server for AI Agents — Industrial AI Nexus
  • September 2025 · Houston, TX Industrial AI Agents and Agent-as-a-Service — Houston Data and AI
  • June 2025 · Houston, TX Empowering AI Agents with MCP and A2A Protocols — Houston Data and AI
  • May 2025 · Houston, TX AI Agents and Future — Houston Data and AI
  • December 2024 · Houston, TX STORM AI and UI Agent Systems — Chevron Corporation — Athena Search Seminar
  • October 2024 · Houston, TX Digital Rock Smart Segmentation 2.0 — Shell International E&P — Knowledge Sharing Session
  • April 2024 · Houston, TX 8K-Resolution Segmenting Every Grain — Shell International E&P — Digital Rock Modeling Session
  • June 2023 · Houston, TX Generative Diffusion Models in Action — Shell International E&P — Generative-AI Reading Group
  • June 2021 · Virtual Explore the Epoch of Reionization Using Deep Learning — Statistical Challenges in Modern Astronomy VII
  • June 2020 · Virtual A Deep Learning Approach to Quasar Continuum Prediction — 236th American Astronomical Society Meeting
© 2026 Bin Liu