Software Engineer · Seattle, WA

Pranav Vempati Software Engineer

Software Development Engineer at AWS, with a Master's in Computer Science and a foundation in deep learning research and high-performance computing.

pranav.k.vempati@gmail.com (408) 507-3418 Seattle, Washington
Scroll ↓

Engineering across the full stack of intelligence.

I work at the intersection of distributed systems and machine learning. I'm currently a Software Development Engineer at AWS in Seattle, building backend systems at scale.

Before AWS, my path ran through generative modeling research at Lawrence Livermore National Laboratory, applied ML engineering, and academic computer vision work. I previously served as President of Santa Cruz Artificial Intelligence, a 150+ member ML organization at UC Santa Cruz. I'm currently open to new opportunities in distributed systems, backend infrastructure, and applied machine learning.

03/2025 — Present

Software Development Engineer

Amazon Web Services (AWS) · Seattle, WA
06/2024 — 03/2025

Machine Learning Engineer

CoreData AI · Remote
  • Engineered and deployed AI chatbots and backend models enabling customers to derive actionable insights from their data — improving time-to-decision by 60% and engagement by 40%.
  • Built the backend in Python on AWS Lambda, exposing scalable endpoints via API Gateway for low-latency, serverless inference.
  • Developed domain-specific prompt engineering and response templates (task decomposition, structured outputs, guardrails), improving relevance and consistency while reducing hallucinations.
08/2022 — 01/2023

Data Scientist

Lawrence Livermore National Laboratory · Livermore, CA
  • Implemented Continuous Conditional GANs (CcGANs) in PyTorch as a learned surrogate for spatial laser energy deposition during 3D metal printing — replacing classical melt-pool simulations requiring HPC-scale compute.
  • Modified the CcGAN loss function to incorporate physical constraints and built custom dataset augmentation pipelines, achieving 93% fidelity to ground-truth physical experiments.
  • Contributed to libROM, a C++ Reduced Order Modeling library — authoring a regression test suite, enhancing the CI workflow with GitHub Actions, and incorporating FEM-based simulations.
09/2021 — 08/2022

President — Santa Cruz Artificial Intelligence

UC Santa Cruz, Baskin School of Engineering · Santa Cruz, CA
  • Led a Baskin Engineering–affiliated organization of over 150 members.
  • Prepared and delivered lessons, and mentored members through their projects. (Lecturer, 2018–2022.)
06/2021 — 06/2022

Researcher — Computer Vision Lab

UC Santa Cruz · Santa Cruz, CA
  • Worked with Professor Roberto Manduchi to maintain and benchmark an iOS Swift-based OCR application for the visually impaired, integrating Google ML Kit features.
06/2020 — 09/2020

Software Engineering Intern — End User Computing

VMware · Palo Alto, CA
  • Built an Angular-based project enhancing the renovated Workspace ONE administration console, adding functionality to the Identity Management UI in Angular 8 and Clarity.
08/2019 — 09/2019

Embedded Systems & Deep Learning Intern

ICURO · Santa Clara, CA
  • Delivered a Python OCR-based license-plate recognition system using TensorFlow Lite MobileNet and Darknet YOLOv3, reaching 96% accuracy on a custom dataset.
  • Shipped 8-bit quantized models for offline on-device inference on Edge TPU / Mendel Linux, achieving a 10× reduction in model size.

What I build with.

Languages

C++PythonCJavaTypeScriptJavaScriptKotlinSwiftSQLBash

ML & AI

PyTorchTensorFlowKerasComputer VisionLLMsTransformersGenerative AINLPReinforcement Learning

Systems & Infra

AWSLambdaDynamoDBCloudWatchDockerKubernetesCUDASLURMDistributed SystemsCI/CD

Libraries & Data

NumPyPandasSciPyScikit-LearnOpenCVONNXPlotlyMatplotlibCMake

Foundations.

01/2023 — 06/2024

M.S. in Computer Science — GPA 3.82 / 4.00

University of California, Santa Cruz

MS Project: Evaluating the effectiveness of fairness techniques for decision-tree classifiers. Benchmarked five interventions — including rule-based pruning, Shapley-importance pruning, and direct modification of scikit-learn's Cython decision-tree training to penalize imbalanced splits — reducing false-positive-rate divergence on the maximally divergent subgroup by 7.82% with no degradation in accuracy or ROC AUC.

09/2018 — 12/2021

B.S. in Computer Science — GPA 3.46 / 4.00

University of California, Santa Cruz