Mechanical Engineering · Ph.D. Candidate

Musa Ibne Mannan

I fabricate biosensors, synthesize 2D materials, and design the experiments that tell me why things fail.

Ph.D. candidate at UT Dallas in Manufacturing and Design Innovations, working across hardware, mechanics and data. I built the lab's CVD line from an empty bench, grew monolayer MoS2 in it, and trained a model on 139 of those growths that carried a working recipe into a different reactor on the first attempt. Before that I fabricated electrochemical aptasensors for tRNA fragment biomarkers, and the next device follows the same route. Before graduate school I ran mechanical and piping projects on a condensate refinery in Bangladesh, which is where I learned root-cause analysis under a deadline, then spent two years supporting a California emergency room through COVID, where I built the world's first remote ER scribing workflow from scratch.

Open to device, process development and materials ML roles Dallas, Texas
Musa Ibne Mannan at an optical microscope in the Leem Research Group lab at UT Dallas
6Journal papers and proceedings, 1 under review
850+Engineering students taught or mentored
120+Field design faults found and resolved
~30,000Man-hours saved on refinery projects

Capabilities

What I work on

01

Biosensor fabrication and electrochemistry

Screen-printed carbon electrodes functionalized by oxygen plasma and EDC/NHS chemistry, with BSA blocking for specificity. Probe immobilization and target hybridization verified by cyclic and differential pulse voltammetry. The published device reads tRNA-derived fragments down to 0.4 nM and separates a three-nucleotide mismatch.

SPCEEDC / NHSCV / DPVPoint of care
02

2D material synthesis by CVD

Specified, assembled and commissioned a dual-zone CVD line with hydrogen generation and in-line Raman, then ran 139 MoS2 growths on SiO2/Si from MoO3 and elemental sulfur. Measured the axial thermal gradient at roughly 37 °C per centimeter, which is the reason published recipes so often refuse to transfer.

CVDMoS₂RamanHigh vacuum
03

Machine learning for process and materials

Random Forest and XGBoost trained on physically referenced inputs rather than controller set-points. The MoS2 growth classifier reached 89.3 percent accuracy with complete recall on growth-positive conditions, and ranked calibrated substrate temperature as the most informative variable of the set.

PythonRandom ForestXGBoostFeature importance
04

Mechanical design, FEA and manufacturing

Abaqus thermal and stress modeling, benchmarked against closed-form solutions and against my own MATLAB quadrature code. Hands-on across twin-screw extrusion, FDM and SLA printing, three-axis CNC on a HAAS VF-3, welding, mill and lathe.

AbaqusSolidWorksMastercamMTS 810

Selected work

Projects

Ordered by how recently the work was published or completed.

Zero-shot transfer of MoS2 growth conditions between reactors

CVD recipes for monolayer MoS2 almost never transfer between furnaces, and the usual reason is that the controller set-point is not the temperature the substrate sees. I measured the axial gradient through the growth zone at roughly 37 °C per centimeter, which means a one-centimeter shift in boat position moves the local temperature by more than the entire useful growth window. Across 139 growths I logged a calibrated substrate temperature instead of a set-point, trained a Random Forest on those inputs to 89.3 percent held-out accuracy with complete recall on growth-positive conditions, then applied it to a geometrically different single-zone reactor. It produced MoS2 in the first trials with no further iteration.

Random ForestMoS₂Thermal calibration139 growths

Advanced Materials, under review

Electrochemical detection of tRNA-derived fragment biomarkers

A disposable screen-printed carbon electrode for tRFs, an emerging biomarker implicated in cancer, neurological disorders and infection. The working electrode is functionalized by oxygen plasma and then EDC/NHS chemistry for probe immobilization, with BSA blocking the unmodified surface against non-specific binding. Cyclic and differential pulse voltammetry confirmed each step. The device reached a 0.4 nM limit of detection over a 1 to 100 nM range and distinguished a three-nucleotide mutation, at a cost that suits resource-limited settings.

BiosensorSPCEDPV0.4 nM LOD

Surfaces and Interfaces 83, 108446, 2026

Urea-SCR with air preheating for diesel NOx reduction

A four-stroke single-cylinder water-cooled diesel engine fitted with a mechanical urea injection system, a platinum-rhodium catalyst, and an air preheater driven by exhaust gas heat. Tested across engine speeds and loads, the combined system cut NOx by up to 55 percent at low load and CO by 19 percent, improved brake thermal efficiency and reduced brake-specific fuel consumption with no significant backpressure. Noise fell 21 to 27 percent, which was not the objective but turned out to matter.

Urea-SCREmissionsPt-Rh catalystEngine testing

Heliyon 11(1), e42399, 2025

Ragweed-PLA biocomposite filament for additive manufacturing

Giant ragweed as a new reinforcement for polylactic acid. We compared retting techniques and found hydrogen peroxide gave the best fiber, treated it with maleic anhydride, GLYMO and AMEO silanes to cut its hydrophilicity, then extruded filament at 10 wt% loading to a 1.65 mm diameter and printed coupons to ASTM D638 and D790. Maleic anhydride treatment lifted tensile strength from 35.8 to 46.20 MPa and modulus from 1.6 to 2.3 GPa over neat PLA. Between the silanes, GLYMO beat AMEO by 3.7 percent on ultimate tensile strength and 27 percent on flexural strength.

Twin-screw extrusionFDMSEM / EDAXTGA / DSC / TMA

International Journal of Composite and Constituent Materials 8(2), 2022

Reduced graphene oxide thin films by pulsed laser deposition

Large-area rGO films deposited by PLD and then improved by pulsed laser annealing. Raman at 532 nm resolved the D, G, 2D and D+G bands for defect assessment; XPS and XRD gave sp2/sp3 and carbon-to-oxygen ratios and tracked the two-theta shift confirming reduction. Thickness came from a Woollam M-2000 ellipsometer, carrier type and mobility from van der Pauw Hall measurement, and transport from resistance versus temperature on a PPMS DynaCool between 1.8 and 300 K, modeled against Arrhenius and variable range hopping.

PLDRamanXPS / XRDPPMS DynaCool

Experimental, Texas State University

Complex-geometry part on a three-axis CNC

A mechanically inspired rook machined from a two-inch RenShape polyurethane cube, with toolpaths generated in Mastercam for a 1/2 inch flat endmill, a 1/16 inch flat endmill and a 1/4 inch chamfer mill. Design and toolpath generation took 4 hours 50 minutes, machining 22 hours 40 minutes. The first attempt could not reach a section 0.661 inches from the bottom; I traced it to the seventh machining step and fixed it by recalibrating the toolpaths and the posting options. Probing set X, Y and Z zero and measured tool lengths before the run.

CNCMastercamG-codeWorkpiece probing

Mastercam and HAAS VF-3

Machine gun barrel thermal and stress failure analysis

WWII machine guns firing up to 1,200 rounds per minute built heat faster than the barrel could shed it, which is why crews carried spares. I calculated the firing pressure analytically at roughly 130.54 MPa, then ran transient heat transfer and von Mises stress analysis in Abaqus with plasticity derived from converted true stress-strain data. AISI 4140, AK Steel 316L and ASTM A656 Grade 7 were compared; 4140 showed the lowest peak von Mises stress, and the steepest thermal gradients sat near the breach after 15 seconds of sustained fire.

Transient thermalvon MisesMaterial selection

Abaqus FEA

Energy plan selection in Texas by multi-objective genetic algorithm

ERCOT residential usage data from 2002 to 2022, used to choose an electricity plan against three objectives that pull against each other: cost, renewable share and cancellation fees. The model converged after 101 generations. Supporting analysis covered hourly consumption distributions, a correlation heatmap showing strong coupling between adjacent intervals, and boxplots characterizing variability and outliers.

MATLABMOGAERCOT data

MATLAB, MECH 6372

Contact

Get in touch

Email reaches me fastest. I am based in Dallas and open to relocating for the right role.

Musa Ibne Mannan at home with the flags of Texas and Bangladesh

Authorship

I am also the author of 26 books, published in two countries

Novelist and screenwriter, writing crime fiction in Bangla as Kishor Pasha Imon. Twenty-six books across Bangladesh and India, written alongside the engineering. Keeping a publishing deadline every year is where I learned to finish things, and to make a complicated argument land for a reader who did not come in already convinced.

  • 26Books published
  • 6,946Pages written
  • 2015 to 2024Publishing years
Cover of ছারপোকা by Kishor Pasha Imon
Cover of যে হীরকখণ্ডে ঘুমিয়ে কুকুরদল by Kishor Pasha Imon
Cover of যে রাতে কাক ডেকেছিল by Kishor Pasha Imon
Cover of যুগলবন্দী by Kishor Pasha Imon
Cover of মায়াবনবিহারিণী হরিণী by Kishor Pasha Imon
Cover of ডিসটোপিয়া by Kishor Pasha Imon
Cover of CHECK TITLE by Kishor Pasha Imon
Cover of মিথস্ক্রিয়া by Kishor Pasha Imon
Cover of চাহিদারা হন্যে by Kishor Pasha Imon
Cover of আগুনের দিন শেষ হয়নি by Kishor Pasha Imon
Cover of নাস্তিকে শিকারি by Kishor Pasha Imon
Cover of অন্তহীন কোলাহল by Kishor Pasha Imon
Cover of দাদা ও আমি by Kishor Pasha Imon
Cover of জোৎস্নায় নিঃসঙ্গ অন্ধ সুন্দরী by Kishor Pasha Imon
Cover of আমরা তোমার শান্তিপ্রিয় শান্ত ছেলে by Kishor Pasha Imon
Cover of এইখানে জাদুঘর পাতা আমাদের by Kishor Pasha Imon