My research vision is for manufacturing to keep pace with AI design, so that every material and part that AI designs can be built accurately, quickly and at a cost industry can use.
Process- and cost-aware AI design
Generative AI
Composition-aware toolpaths for graded parts
Bayesian optimization · agentic AI
Transferable AI process optimization
Physics-informed & transfer learning
Adaptive control of composition
Learned control · multi-modal sensing
Affordable, certifiable graded metal parts
Aerospace · nuclear · defence
AI is speeding up the design of alloys, architectures and parts, and many of the most promising designs, especially functionally graded parts, are best made by additive manufacturing (AM). Manufacturing has become the bottleneck: AI designs rarely consider what can be built or at what cost, and AM cannot yet build them accurately or quickly enough.
My lab, the Additive Innovation and Artificial Intelligence (AI²) Lab, will work on both sides: building manufacturability and cost into AI design, and making AM accurate and reliable enough to build what AI designs.
What I have built so far, and the part of my lab's plan each result supports. Published marks peer-reviewed results; Ongoing marks current, unpublished work.
Ongoing at MIT Design, fabrication and testing of a component-level, functionally graded gas-turbine blisk (DARPA METALS program), using a dilution- and thermal-aware planner that co-optimizes toolpath and composition for multi-material DED.
AI: bilevel optimization
Published Bi-metallic 316L/17-4PH heterostructured steels with 27.3% higher tensile ductility than 316L, explained by a confined-necking co-deformation mechanism (Acta Materialia 2026). Multi-metal parts with magnetic, data-encoding, sealing and mechanical functions, with two patent applications (MetalMat 2026). Process windows for multi-track and multi-layer builds (Additive Manufacturing 2025).
Ongoing Radial bi-metallic materials that overcome the strength–ductility trade-off under ultra-high-speed impact (manuscript submitted).
AI: machine learning + genetic-algorithm inverse process optimization (AIDED)
Published From target mechanical properties directly to 3D microstructures, with a 3D convolutional neural-network surrogate and a genetic algorithm (Materials Today 2023).
AI: 3D-CNN surrogate + genetic algorithm
Published Invited review of microstructure and composition engineering by DED, including in-situ alloying and graded builds (Additive Manufacturing 2026); co-author on active learning for NiCoCr alloy discovery in AM (npj Advanced Manufacturing 2026).
Ongoing at MIT A hybrid laser powder bed fusion + inkjet process for voxel-level, in-situ alloying of functionally graded materials.
AI: active learning
Published Mechanism and quantification of melt-pool morphology in laser DED, combining CFD with in-situ high-speed infrared imaging (J. Manufacturing Processes 2025, co-first author); co-author on ML-aided keyhole simulation (IJHMT 2025).
Ongoing Deep-learning closed-loop control of DED (manuscript in preparation; patent application in preparation). Magnetic-field-assisted DED of 316L and IN718 observed by in-situ synchrotron X-ray imaging with Prof. Peter D. Lee at UCL (manuscript in preparation).
AI: automated image analysis & learned control
1.1 Process- and cost-aware AI design
1.2 Composition-aware toolpath planning
2.1 Transferable AI process models
2.2 Adaptive closed-loop control for graded AM