1-3 mm vs <200 um
LMD focus size compared with SLM in Sharma et al.
Our lattice-structure LMD paper contrasts LMD's larger focus size with SLM's smaller focus, which is why the process design logic is different.
Sharma et al., Procedia CIRP 2018 Personal profile Exafuse-linked Industrial AI
About Manish Sharma
Building industrial AI connected to evidence.
I am Manish Sharma. I work on industrial AI and decision systems for engineering processes. My deepest public work is in AI for Laser Metal Deposition and Directed Energy Deposition at Exafuse, where process monitoring, robotics, repair decisions, RFQ quality, and physical inspection all meet.
The question that connects my public work is not simply whether an AI model can produce an answer. It is whether the decision can be traced to the right signals, assumptions, engineering context, and verification evidence.
Manish Sharma
AI for Laser Metal Deposition
Since
2018
Laser Metal Deposition work
Public role
AI + R&D
Exafuse, Germany
Research
Public profile material
Ruhr University Bochum
Focus
Industrial AI
Decision systems, monitoring, LMD/DED
Current work
Where my work sits
I work in the middle of process development, monitoring, robotics, and AI. The aim is to make industrial decisions clearer without pretending software replaces engineering evidence.
Public profile materials describe a background that combines DED/LMD process development, machine vision, neural-network-based feature extraction, robotic toolpath refinement, and traceable build documentation. The practical thread is process stability, repeatability, and the evidence needed after a decision leaves the model.
This site is where I turn that work into public notes, frameworks, and small tools. It is meant for engineers, students, buyers, developers, and anyone who needs clearer language around LMD, DED, laser cladding, monitoring, repairability, and RFQ preparation.
Public profile details should be verified against LinkedIn, a current CV, or official event material where required. This site avoids confidential employer or customer data; the public technical content is educational and decision-support oriented.
Working numbers
The scale I keep in mind
I keep a 500-record LMD/DED reference map for orientation, but the numbers below are only used where a checked source or public profile material supports them.
1070 nm / 450 W / 2 mm
Parameters from Manish Sharma's 2018 LMD paper
The published experiment used a ytterbium fiber laser, 316L powder, a 2 mm focus size, and a 3 mm substrate for columnar built-up lattice experiments.
Sharma et al., Procedia CIRP 201845-90 um
316L powder size in the 2018 lattice experiment
I keep details like this visible because they are more useful than a generic portfolio claim.
Sharma et al., Procedia CIRP 20185-10 mm / >95%
BreitbahnDED project targets
From my public profile material: broad-track DED goals include rotating multi-spot optics, 5-10 mm wide tracks, multimodal monitoring, layer-to-layer control, and a >95% powder-utilization target. This is a project target, not a published result claim.
Manish Sharma public profile materialFocus areas
What I work on
Industrial AI and decision support
Decision-support systems that keep signals, assumptions, risks, recommendations, and verification evidence traceable.
Process monitoring and machine vision
Vision-based monitoring, melt-pool and bead-geometry indicators, anomaly signals, and data pipelines that make process behavior easier to review.
Robotic industrial workflows
Robotic DED toolpaths, KUKA workflows, slicer post-processing, ROS2 integration, sensor and camera control, and real-time system interfaces.
LMD/DED process understanding
Process windows, repair and cladding workflows, traceability packs, inspection-ready documentation, and repeatability.
Inspection-aware AI
AI outputs treated as decision support, with physical inspection and engineering evidence kept visible where risk requires it.
RFQ and engineering information structure
Frameworks, schemas, prompts, and public notes that help people separate known facts, missing information, assumptions, and risks.
Exafuse proof context
The public industrial work behind the themes
The site should not just use words like AI, monitoring, repair, and RFQ. It should point to the industrial situations where those ideas actually matter.
CS15
Duisburg Bridge Components
This is the strongest public anchor for my interest in AI-assisted process understanding without replacing inspection evidence.
CS01
Forging Hammer Repair
This maps directly to the LMD Repairability Index and the Quality Evidence Ladder.
CS10
Nobufil Extrusion Screw Repair
This is a clean RFQ-intelligence example: damage boundary, lead time, machining route and inspection context decide the recommendation.
CS13
130 mm Build-and-Coat Drill
This supports the site's build-and-coat logic: geometry and surface function should be evaluated as one workflow.
Field principles
What makes the work personal
These are the instincts behind the frameworks: stay practical, keep inspection visible, and be skeptical of overconfident AI claims.
Height is not a cosmetic variable.
In DED/LMD, stand-off distance and layer-height errors add up quickly. I treat height sensing, toolpath correction, and post-machining allowance as part of the process plan, not a late fix.
A camera is not a certificate.
Coaxial vision, pyrometry, and melt-pool features are useful because they tell us what the process was doing. They only become strong evidence when they are connected to inspection results.
RFQs fail when they hide risk.
A weak request is not always a short request. The real problem is missing material grade, damage depth, tolerance, operating conditions, or inspection criteria.
Traceability is a product feature.
For industrial DED/LMD, parameter logs, change control, build reports, inspection reports, and deviation tracking are part of the work, not paperwork at the end.
Experience
Career timeline
Jan 2024 - Present
Exafuse GmbH, Bochum
Head of AI and R&D
Technology owner for monitoring and control concepts for industrial DED/LMD cells, including robotic and large-format systems, sensor/camera stacks, calibration routines, traceability, and inspection-ready reporting.
Jan 2020 - Dec 2023
Exafuse GmbH
Machine Learning and Systems Engineer
Built machine-vision monitoring systems for LMD, developed neural-network models for feature extraction and quality indicators, integrated model outputs into automation interfaces, and supported toolpath refinement.
Dec 2017 - Jun 2021
Ruhr University Bochum, Chair of Applied Laser Technology
Research Assistant / Wissenschaftlicher Mitarbeiter
Worked on vision systems and machine-learning approaches for Laser Metal Deposition, including lattice-structure deposition experiments and process understanding.
Technical stack
Methods, systems, and materials
A compact map of the technical vocabulary that shows up repeatedly in my public work.
Directed Energy DepositionLaser Metal DepositionDED-LB/MRepair and claddingThin-wall buildsCoaxial visionPyrometryHeight sensingROS2KUKA workflowsPythonC++MATLABProcess windowsTraceability packsInspection evidence
Education
Academic foundation
External PhD Researcher
Ruhr University Bochum, Chair of Applied Laser Technology
Vision-based process control for Laser Metal Deposition with Artificial Intelligence. Mar 2020 - Present.
M.Sc. Lasers and Photonics
Ruhr University Bochum
Faculty Prize / Best Student, grade 97%. Apr 2017 - Jan 2020.
B.Tech. Electrical Engineering
Rajasthan Technical University, India
Gold Medal, grade 87%. May 2012 - Jun 2016.
Public work
Selected talks, publication, and awards
Public work
- Invited Speaker / Expert - Outokumpu Metal Powder Event, Krefeld, Germany, Sep 25, 2025.
- Speaker - Data Science Ruhrgebiet 2021, Bochum.
- Publication - Laser metal deposition of lattice structures by columnar built-up, CIRP LANE 2018.
- M.Sc. thesis - Machine learning approaches in laser metal deposition of lattice structures, 2019.
Awards and signals
- Material Horizon Prize - Best presentation at CIRP LANE 2018.
- DAAD Graduation Scholarship for International Students, 2019.
- C++ Nanodegree - Udacity, 2022.
- TOEFL iBT 112/120; German B1.
Human note
Why this site exists
I’m interested in how AI can improve technical work without blurring the line between a useful process signal and validated engineering evidence.
The goal is to make complex LMD knowledge easier to structure: for buyers preparing RFQs, engineers reviewing risk, and developers building technical tools. This site stays public, inspection-aware, and careful about confidential or employer-owned information.
Identity path
Continue through the public identity layer
Industrial AI for Decisions You Can VerifyThe operating thesis behind the public lab.Open →Manish Sharma - Industrial AI & Decision SystemsThe canonical public identity page.Open →LMD / DED Domain HubThe established public proof domain.Open →Work & ProofPublic frameworks, tools, evidence, and artifacts.Open →ExafuseIndustrial LMD/SLM services, case studies, and RFQ context.Open →Full Site MapEvery page and public machine-readable asset in one linked map.Open →