PREMITH KUMAR CHILUKURI Senior AI Engineer & Researcher Contact: reveal email in the website Contact section GitHub: https://github.com/chpk Google Scholar: https://scholar.google.com/citations?user=tVK5rDMAAAAJ&hl=en LinkedIn: https://www.linkedin.com/in/cpremithkumar/ EXPERIENCE ABB | Senior AI Engineer | 2024 — PRESENT Senior AI Engineer · Mar 2025–present / AI Intern · Jan–Mar 2025 / Robotics Intern · May–Dec 2024 - Enterprise agent orchestration: Architected a natural-language automation platform with a hierarchical supervisor–specialist runtime, parallel DAG execution, stateless container workers, and Kubernetes scaling. - Knowledge at industrial scale: Built an agentic multimodal RAG platform spanning a 400K+ document corpus, a 419K-entity knowledge graph, and 1.66M relations, using PostgreSQL, Apache AGE, pgvector, query expansion, and relevance grading. - Long-running agents, under control: Engineered persisted human approval checkpoints, automatic context compression, token budgets, dynamic skill loading, sandboxed Python, and end-to-end tool-call tracing. - Connected tools and usable outputs: Implemented MCP integrations, OAuth connectors, structured SAP analytics, and versioned document generation. Served the platform through FastAPI and deployed it in Docker with multi-provider inference. - Research into engineering practice: Contributed to MiRAGE and Power Circuit AI. Earlier ABB work covered evaluation across 500+ queries with 10+ metrics, motor search over 500+ specifications, and a 6D pose pipeline reporting 5 mm MSSD at 0.6 s inference. Virginia Tech | Graduate Research Assistant | 2023 — 2024 March 2023 – May 2024 - Manufacturing-aware generation: Co-authored guided voxel diffusion research for porous microbial fuel-cell anodes: 64³ voxel designs with a learned minimum-feature-size surrogate guiding the later denoising stages. - 3D perception and reconstruction: Investigated single-view NeRF synthesis, diffusion-based reconstruction, quantized language models, and LoRA adaptation for 3D object generation and pose-estimation workflows. - Retrieval and agent systems: Built database-connected agents using LangGraph, Tree-of-Thought, and ReAct; developed hybrid retrieval and HyDE-based RAG for business and process-optimization applications. - Research made explorable: Developed an augmented-reality visualization of 186 neural-network results across 85 UCR time-series datasets for the World Map of AI research project. Supervue AI | Lead Machine Learning Engineer | 2020 — 2023 Lead ML Engineer · Feb 2021–Jan 2023 / ML Engineer · Jan 2020–Feb 2021 - Computer vision in production: Led the AI application team and developed a facial-recognition API adopted by 50+ clients, alongside real-time person detection and tracking used in nationwide examinations and other deployments. - Systems built for scale: Scaled AI applications to 30K users with microservices; the profile reports 99.1% uptime for the monitoring application and 10% lower deployment costs. - Document and speech intelligence: Developed distributed OCR processing for a reported 100M+ documents, improved speech transcription accuracy by 25.2%, and advanced intent classification with fine-tuned language models. - Applied invention: Developed a social-distance monitoring method associated with application 202041034063, as well as generative image perturbation and privacy-oriented facial-image processing. Samsung R&D | Research Collaborator | 2021 — 2022 October 2021 – August 2022 · PRISM collaboration with Andhra University - Real-time visual reconstruction: Engineered a multi-threaded inpainting algorithm using lightweight feature transformation, search modules, and caching; the profile reports 31 FPS and a 15% reduction in distortion artifacts. - From method to invention: Contributed the nN-TFT approach for edge deployment, associated with the video-inpainting patent application reference 202241043300. - Industry–academic collaboration: Conducted the research through Samsung R&D Institute Bangalore’s PRISM collaboration with Andhra University. Zebi | Machine Learning Engineer | 2018 — 2019 December 2018 – December 2019 - Faster identity workflows: Helped build a blockchain-assisted check-in system with biometric verification and ID-card OCR APIs, reporting a 40% reduction in hotel check-in time. - Operational machine learning: Built ticket summarization, build-failure prediction, and time-series alerting; the profile reports 0.95 recall for build-failure prediction and a 25% downtime reduction across 295+ servers. PROJECTS MiRAGE A collaborative agent framework that turns complex documents into verified, multimodal, multihop evaluation datasets. Equal-contribution co-author of the MiRAGE research with Chandan Kumar Sahu and Matthew Hetrich. The public framework and paper document the team’s approach to generating domain-specific evaluation datasets. Evaluated on finance, regulation, quantitative biology, and journalism. The paper reports different reasoning complexity by domain; journalism remains closer to single-hop. Visual grounding is an explicitly acknowledged limitation. Research paper: https://arxiv.org/html/2601.15487v1 Source code: https://github.com/ChandankSahu/MiRAGE Python package: https://pypi.org/project/mirage-benchmark/ Power Circuit AI An autonomous multi-agent workflow connects natural-language specifications with component selection, schematics, PCB layouts, and iterative rule checks. Co-authored the Power Circuit AI work with Chandan Kumar Sahu, Matthew Hetrich, and Sneha Narasimhan at ABB US Corporate Research Center. The published case study demonstrates logical schematic-to-PCB connectivity for a variable-frequency drive with a braking chopper. Physical layout, thermal behavior, EMI, and manufacturing readiness still require expert engineering review. Preprint: https://engrxiv.org/preprint/view/6706/version/8726 IEEE publication DOI: https://doi.org/10.1109/IDCD69431.2026.11519438 Guided Voxel Diffusion A 3D denoising diffusion model, guided by a manufacturability surrogate, generates porous microbial fuel-cell anode structures. Contributed to the guided voxel diffusion research with Binyang Song, SungKu Kang, and Ran Jin, connecting 3D generative modeling with manufacturing-aware optimization. The study reports normalized minimum feature size of 0.63 for guided designs versus 0.54 for the unguided model. It also describes a quality–diversity trade-off; improved manufacturability is not the same as universal improvement across every metric. Guided diffusion study: https://vtechworks.lib.vt.edu/bitstreams/ccf10eb3-a629-474e-9772-f7e1f324b349/download MSEC publication: https://doi.org/10.1115/MSEC2024-125075 Enterprise Agent Orchestration A production-oriented orchestration platform where agents plan, delegate, use tools, and resume complex work with human review. Architected and engineered the orchestration platform, runtime, context-management layer, dynamic skills, and data-access integrations in my current ABB role. The architecture supports long-running work, controlled tool execution, observable reasoning workflows, and human-in-the-loop review. These implementation details are drawn from my professional profile; no public product demo or benchmark is claimed. Professional profile: https://www.linkedin.com/in/cpremithkumar/ Multimodal Enterprise RAG Hybrid retrieval joins a large enterprise knowledge graph with multimodal documents, structured analytics, and grounded agent workflows. Built and deployed the agentic multimodal RAG platform, knowledge-graph infrastructure, and data-analysis/document workflows described in my ABB experience. The profile reports a corpus of more than 400,000 documents. The graph, document count, and relation count measure different parts of the system; they are not independent measures of answer quality. Professional profile: https://www.linkedin.com/in/cpremithkumar/ SKILLS Production AI & deployment: Python, FastAPI, Docker, Kubernetes, AWS / Azure, CI/CD, Inference optimization Agentic AI & reasoning: Multi-agent systems, MCP, LangGraph, LangChain, Context engineering, Evaluation Retrieval & knowledge systems: Multimodal RAG, Knowledge graphs, pgvector, Apache AGE, PostgreSQL, Document parsing Generative research & vision: PyTorch, Diffusion models, LoRA, NeRF, 3D reconstruction, OpenCV, TensorRT / ONNX RESEARCH DATA OpenAlex: https://openalex.org/A5015604978 Snapshot: 2026-09-28T19:58:03.524431+00:00 Citation links: 20; Countries: CN, ES, GB, IN, PT, US Affiliations reflect citing authors, not institutional endorsements. Publications and invention references: see the Research section of the HTML profile, including all source links and status notes. EDUCATION Virginia Tech | Master’s degree · Computer Science Blacksburg, Virginia BML Munjal University | Bachelor of Technology · Computer Science Gurugram, India SERVICE & OUTREACH 20+ research articles reviewed Reviewed 20+ research articles for IEEE Access, Engineering Applications of Artificial Intelligence, and Springer Nature’s Signal, Image and Video Processing. AI for Grid Resilience and Security Challenge Served on the review panel for the AI for Grid Resilience and Security Challenge innovation spotlight sessions at TechConnect World Innovation Conference and Expo 2026. https://www.techconnectworld.com/World2026/AI-for-Grid-Resilience-and-Security-Challenge/ ABB US Accelerator Served as a key member of the jury panel for the 2026 ABB US Accelerator event, contributing technical judgment on emerging innovations. INFORMS QSR university hackathon Developed and helped organize a university hackathon for the INFORMS Quality, Statistics and Reliability (QSR) 2024 challenge. Concepts of Manufacturing & Industrial Internet Presented Concepts of Manufacturing & Industrial Internet for Virginia Tech’s Pathways for Future Engineers initiative in 2023, introducing students to manufacturing and connected industrial systems. https://eng.vt.edu/ceed/ceed-pre-college-programs/pathways.html GOOGLE SCHOLAR SNAPSHOT 39 citations; h-index 4; i10-index 2. From profile-owner supplied screenshot, recorded September 28, 2026. Separate from the OpenAlex geographic dataset. PUBLICATIONS 2026 | Power Circuit AI: Designing Power Electronic Circuits for Motor Drives with Generative Artificial Intelligence IEEE International Conference on Industrial Digitalization (IDCD) | Conference paper An agent workflow for power-electronics design, from specifications and components through schematics, connectivity and design checks. https://doi.org/10.1109/IDCD69431.2026.11519438 2026 | MiRAGE: A Multiagent Framework for Generating Multimodal Multihop Question-Answer Dataset for RAG Evaluation arXiv · 2601.15487 | Preprint Equal-contribution co-author. A multiagent method for generating and verifying questions that require retrieval and reasoning across multimodal evidence. https://arxiv.org/abs/2601.15487 2025 | High-Quality Dataset-Sharing and Trade Based on A Performance-Oriented Directed Graph Neural Network IEEE Transactions on Automation Science and Engineering | Journal article Performance-oriented representations and a directed graph neural network help rank datasets for manufacturing tasks while preserving private information. https://doi.org/10.1109/TASE.2025.3561081 2025 | High-Quality Dataset-Sharing Based on Directed Graph Neural Network IEEE CASE · profile record | Profile-listed conference contribution A related contribution on choosing manufacturing datasets with directed graph learning. https://www.linkedin.com/in/cpremithkumar/details/publications/ 2025 | High-Quality Privacy Preserving Text Data Sharing for Manufacturing Machine Learning Data Market IISE Annual Conference and Expo 2025 | Conference contribution Privacy-preserving text data sharing for manufacturing machine-learning applications. https://www.proceedings.com/content/081/081540webtoc.pdf 2024 | Generative Design for Manufacturing: Integrating Generation with Optimization Using a Guided Voxel Diffusion Model OpenReview · manuscript record | Research manuscript Integrating generation with manufacturability guidance for porous three-dimensional designs. https://vtechworks.lib.vt.edu/items/469b7241-3222-4992-9b92-578d3a43208e 2024 | Generating Optimized 3D Designs for Manufacturing Using a Guided Voxel Diffusion Model ASME Manufacturing Science and Engineering Conference (MSEC) | Conference paper Guided voxel diffusion generates porous microbial fuel-cell anode structures while targeting minimum feature size. https://doi.org/10.1115/MSEC2024-125075 2021 | Attention Based Multi-Patched 3D-CNNs with Hybrid Fusion Architecture for Reducing False Positives during Lung Nodule Detection Journal of Computer and Communications | Journal article Combining multi-patch three-dimensional representations with attention to reduce false positives in lung nodule detection. https://doi.org/10.4236/jcc.2021.94001 2021 | l, r-Stitch Unit: Encoder-Decoder-CNN Based Image-Mosaicing Mechanism for Stitching Non-Homogeneous Image Sequences IEEE Access | Journal article Learning to stitch image sequences with mismatched visual characteristics using an encoder–decoder CNN. https://doi.org/10.1109/ACCESS.2021.3052474 2020 | AMMDAS: Multi-Modular Generative Masks Processing Architecture With Adaptive Wide Field-of-View Modeling Strategy IEEE Access | Journal article A modular perception architecture combining generative masks and adaptive wide-field-of-view modeling. https://doi.org/10.1109/ACCESS.2020.3033537 PATENTS & INVENTION REFERENCES Identifier status: supplied references, not uniformly confirmed grants. A Novel Architecture To Perform Fast (Real-Time) Video Inpainting 202241043300 | Computer vision ARCHITECT: Automated Robust CAD‐based Hierarchical Intelligent Engineering, Neural Evaluation, Testing, and Assembly Toolkit P250309US01 | Engineering AI Agentic Architecture for Automated Assembly of Parametric CAD Models Using Advanced Reasoning Paradigms with Generative AI P250244US01 | Engineering AI Agentic Swarm‑Driven Autonomous Data Science System for Real‑Time Industrial IoT Analytics P250351US01 | Industrial intelligence Autonomous Generation and Advisory of 3D CAD Assemblies from Multimodal Engineering Data P250513US01 | Engineering AI Dynamic Auto-calibration System for Spatial Depth Analysis at Crowded Places 202041034063 | Computer vision Generative Adversarial Agentic Architecture for Automated Parametric CAD Modeling and Multi-Physics Simulation Optimization Using Advanced Reasoning Paradigms P240832US01 | Engineering AI Hypothetical Product Embedding for Industrial Devices P250510US01 | Industrial intelligence Methods for Building Domain Knowledge Databases from Industrial Data Sources P250509US01 | Industrial intelligence Multi-Stage Visual Language Model Enabled Agentic Redline Interpretation and Parametric CAD System for Automated Design and Manufacturing Integration P250273US01 | Engineering AI OPTIC: Orchestrated Perceptual Thought Iteration for Cognitive visual analysis P240831US01 | Computer vision VLM-based Agentic Process for Creation of Parametric 3D CAD Models P240846US01 | Engineering AI