Mehnaj Jimy

Chemical & Molecular Engineering · Nanotechnology

Stony Brook UniversityB.E. expected May 2028

Undergraduate Researcher · Lee Research Group

My research is in thin-film processing, from spin-coating recipe development to ALD and vapor-phase infiltration on polymer films. Outside the lab, I build process monitoring and yield analysis projects with public semiconductor fab data.

Seeking Summer 2027 internships in semiconductor process and yield engineering.

Resume PDF
CV Portfolio PDF LinkedInGitHub

Research

ALD Growth on Polymer Materials

I work in a two-person polyurethane thin-film team. We develop spin-coating recipes with designed experiments and Bayesian optimization, run ALD and vapor-phase infiltration on the films, and characterize them with ellipsometry, in-situ QCM, FTIR, XRD, SEM, and STEM.

Research overview
ALD instrument with sample vials and tools on top; background edited
ALD system used in my research · edited photo

Engineering projects

WM-811K Wafer Map Yield Analysis

Independent project · Public WM-811K wafer maps

I ranked defect patterns by the dies they cost across 172,950 labeled wafer maps. Edge-Ring and Edge-Loc account for 47% of the excess yield loss.

All 8 defect patterns cluster within lots, and 502 of 7,025 lots are excursions, which points to lot-level causes rather than isolated wafer events.

Case study GitHub
Bar chart of extra dies lost per 1,000 labeled wafers by defect pattern; Edge-Ring and Edge-Loc lead, reaching 47% of cumulative excess loss.
Excess die loss by pattern against clean wafers of the same map size · WM-811K

PVD Process Monitoring & Virtual Metrology

Independent project · Public Infineon datasets

I flagged multivariate process excursions in 97-channel equipment sensor data from Infineon’s public AlCu and WTi PVD release, then ranked the channels behind each one for root-cause review.

A machine-learning virtual metrology model predicted all 17 released film-thickness targets and cut prediction error 41–43% against a linear regression baseline.

Ordinary-split aggregate R²: AlCu 0.622 · WTi 0.604

Case study GitHub
Two questions from recorded process dataRecorded AlCu and WTi process data branches into statistical monitoring to review runs, and virtual metrology to predict 17 targets.Recordedprocess dataAlCu / WTiWhich runs need review?Statistical monitoringCan the targetsbe predicted?17-target virtual metrologyTwo questions from recorded process dataRecorded AlCu and WTi process data has two separate uses: statistical monitoring to review runs and virtual metrology to predict 17 targets.Recorded process dataAlCu / WTiWhich runs need review?Statistical monitoringCan the targets be predicted?17-target virtual metrology

Physics-Based ALD Reactor Model

Active independent project

My Python model follows precursor transport and surface reactions through an ALD cycle. On a synthetic recipe, the well-mixed model predicts 97.8% surface completion everywhere, while the spatial model falls to 91.1% at the outlet. The well-mixed assumption hides under-dosing near the outlet.

Working note GitHub
Well-mixed and spatial ALD surface completionSaved synthetic model output at the end of the A pulse. The well-mixed prediction is 97.8 percent throughout the reactor. The spatial prediction falls from approximately 100 percent near the inlet to 91.1 percent at the outlet. The vertical scale spans 90 to 100 percent.Surface completion (%)90100Well-mixed97.8%Spatial · 91.1%InletOutletWell-mixed and spatial ALD surface completionSaved synthetic model output at the end of the A pulse. The well-mixed prediction is 97.8 percent throughout the reactor. The spatial prediction falls from approximately 100 percent near the inlet to 91.1 percent at the outlet. The vertical scale spans 90 to 100 percent.Surface completion (%)90100Well-mixed97.8%Spatial · 91.1%InletOutlet
Synthetic model output · not experimental data

ColdPlateTwin: Water-Cooled Testbed Design

Active independent project · Preliminary CAD and synthetic simulations

I’m designing a water-cooled testbed with interchangeable flow inserts. I use CAD and Python to examine heat balance and pressure losses before physical testing.

Case study GitHub
Oblique view of the preliminary ColdPlateTwin CAD assembly
Preliminary CAD · not built hardware

Heat and Mass Exchange Simulator

I modeled a counter-current heat exchanger and membrane mass transfer in Python. Changing flow rates and transfer area lets me compare outlet conditions, and energy and solute balances check the calculations.

Project note GitHub

Batch Reactor Simulator

I solve batch reaction kinetics in Python to follow concentration and conversion over time. Each case is checked against its analytical solution, so the numerical error is visible alongside the result.

Project note GitHub