Aamer Kazi
FEA and Deep Learning at Align Technology
Profile

About Me

Research engineer combining computational mechanics, simulation, and machine learning to solve problems in manufacturing and healthcare. My journey started with numerical modeling of rock drilling and metal cutting processes, and has since moved into production deep learning systems, geometric representation learning on 3D meshes, and LLM-powered agents. I enjoy the space where physics and data meet, and I've found that letting the problem structure guide the model architecture tends to be what makes things work.

My technical range covers deep learning (Transformers, autoencoders, representation learning), generative AI (agents, VAEs), and classical ML, alongside high-fidelity nonlinear finite element simulation and constitutive modeling in ABAQUS.

At Align Technology, I build and validate production models serving millions of patients under FDA oversight. That includes writing AI evaluation reports with acceptance criteria, fairness audits across 13+ demographic and clinical dimensions, and full V&V documentation for deployment.

Resume (PDF)

Skills & Expertise

Machine Learning & AI
PyTorchTensorFlowTransformers / HuggingFacescikit-learnXGBoost / LightGBMDiffusionNetOptunaLangGraphMLflowClearMLPhysics-Informed ML
Simulation & Computational Mechanics
ABAQUS (Explicit & Standard)Nonlinear FEAConstitutive ModelingDrucker–Prager / Johnson–CookDamage & Fracture MechanicsExperimental Validation & V&VManufacturing Process Simulation
Data & Scientific Computing
PythonMATLABNumPy / SciPyPandasOpen3D / TrimeshGeometric PriorsConstraint-Aware Optimization
Engineering & Infrastructure
GitDockerAWS BedrockVertex AIFDA Regulatory (ML in Production)CI/CD for ML Pipelines

Work Experience

Align Technology
R&D Engineer III, Modeling and Simulation
San Jose, CADec 2023 — Present
ML / Deep Learning

Deep Learning Enhanced Treatment Planning for the Invisalign System

  • Developed production ML/DL models for orthodontic treatment planning, iterating from multivariable regression to a Transformer architecture (multi-head attention, PyTorch) that contributed to a 20% increase in clinical efficacy for millions of patients annually
  • Led verification & validation for all model generations under FDA regulatory oversight; performed data-slice analysis across 13+ demographic and clinical dimensions, physics-based simulation validation, and detailed AI model reporting ensuring no subpopulation degradation
  • Generated surrogate training labels via counterfactual analysis through the production Transformer across 30,000+ cases, then framed attachment placement as a Learning-to-Rank problem (LightGBM LambdaRank); extracted interpretable clinical rules validated by domain experts and deployed to production processing 1,000+ cases/week
  • Led an end-to-end clinical study (APAC region) assembling a cross-functional team of clinical experts, software developers, and CAD designers
PyTorchTransformersLightGBMFDA V&V
Agentic AI

LLM-based Dental Treatment Assistant & Evaluation Framework

  • Architected a LangChain ReAct agent integrating LLMs (Claude, Llama, Gemini) over AWS Bedrock for natural-language querying and modification of orthodontic treatment data, with a 3-category evaluation framework (classification accuracy, calculation tolerance, LLM-as-Judge format compliance)
  • Built a middleware layer with query sanitization and complexity classification for dynamic per-request model routing; currently being tested as a case screening tool for clinical studies
LangChainAWS BedrockLLM-as-JudgePython
Geometric Deep Learning

3D Shape Representation Learning for Orthodontic Attachments

  • Built a DiffusionNet-based mesh autoencoder encoding variable-topology 3D meshes into 8-12D latent vectors at sub-millimeter fidelity; conducted systematic AE vs beta-VAE comparison across multiple quantitative metrics (reconstruction fidelity, generative specificity, disentanglement, interpolation smoothness)
  • Designed embeddings as drop-in geometric features for the downstream treatment planning Transformer, replacing categorical attachment labels with continuous geometric descriptors
DiffusionNetPyTorchAERepresentation Learning
Texas A&M University
Graduate Research Assistant
College Station, TXAug 2018 — Dec 2023
U.S. DOE

Finite Element Modeling of Rock Drilling Processes

  • Developed a finite element model in ABAQUS for PDC rock cutting simulation using the Drucker-Prager yield criterion with damage evolution to capture failure of granite under varying crack density, orientation, and confining pressure (0-100 MPa)
  • Calibrated constitutive model parameters through uniaxial and diametrical compression testing; cross-referenced cutting forces predicted by the finite element simulation with experimental results for validation
  • Built a MATLAB-based node separation algorithm generating random pre-existing cracks between elements based on configurable crack density parameters, enabling parametric study of crack effects on cutting forces and mechanism
  • Constructed a laboratory-scale rock drilling setup on a CNC mill, demonstrating 30% reduction in cutting and thrust forces from plasma-induced pre-cracking
ABAQUSDrucker-PragerMATLABV&V
Experiment
Simulation
Honeywell

Machining of Additively Manufactured Metals

  • Developed a 2D finite element model in ABAQUS with Johnson-Cook plasticity and damage criteria to simulate serrated chip formation during turning of Ti-6Al-4V; investigated the effect of fracture energy on chip morphology
  • Analyzed machinability of additively manufactured Ti-6Al-4V, A205, and 17-4 SS through a full factorial DOE with SEM-EDS microstructural characterization
ABAQUSJohnson-CookSEM-EDSDOE

Personal Projects

Generative AI

RecipeGPT

A 134M-parameter Transformer trained from scratch on 2.3M recipes.

Custom GPT-style decoder-only transformer (12 layers, 4 heads, 768-dim) built and trained from scratch in PyTorch — no HuggingFace, no pretrained weights — on 398M tokens of recipe text, reaching a validation loss of 1.75. Generates a quantified ingredient list and method from a handful of ingredients. The Lab page opens the model up: a recorded generation with the distribution behind every token, and all 48 attention heads labelled by their measured behaviour.

PyTorchFrom ScratchTransformersPython
␣start␣titleChocChipCook␣ingr2
Causal self-attention · Layer 0, Head 0

Education

Texas A&M University
Ph.D., Mechanical Engineering · GPA: 3.74 / 4.0
College Station, USAAug 2018 — Sep 2023
Vellore Institute of Technology
B.Tech., Mechanical Engineering · GPA: 9.17 / 10
Vellore, IndiaAug 2014 — Jun 2018

Contact

Emailaamerk4716@gmail.comLinkedInlinkedin.com/in/aamerk4716/