Hello, I'm

Ishani Kohli

Software Engineer · AI/ML · Full-Stack

M.S. Computer Engineering · Virginia Tech, 2026

PythonTypeScriptReactPyTorchGCPKubernetes
42%User growth
30%Fewer vulnerabilities
10+Production APIs
Scroll ↓

I am a software engineer and AI/ML engineer with a passion for building systems that are both robust and user-centered.

My interests lie in applied machine learning, full-stack development, and cloud-native engineering — from shipping production microservices on GCP/Kubernetes to researching how large language models behave under adversarial stress. I enjoy projects that combine strong engineering fundamentals with thoughtful product design.

During my career

I've worked across the full SDLC — building a code-quality platform at Persistent Systems that reduced vulnerabilities 30%, shipping RESTful APIs and CI/CD pipelines on GCP, and leading frontend modernization of a legacy exam platform at Virginia Tech that grew the user base 42% while improving accessibility for 2,000+ students.

Now

I recently completed my M.S. in Computer Engineering at Virginia Tech, focusing on Software & Machine Intelligence. I'm looking to combine full-stack engineering, applied ML, and cloud infrastructure to build products that solve real problems at scale.

Recent projects

All projects →
  • Web Development

    Smart Resumes — Full-Stack Resume Builder

    Developed a full-stack resume builder with multi-template live A4 preview and one-click PDF export, solving cross-component state with Redux and debounced Firestore auto-save; deployed via Vercel CI/C…

    ReactReduxFirebaseVercel
  • Machine Learning

    Investigating Antifragility in Large Language Models

    Investigated the robustness of LLMs adapted for image classification with LIFT, measuring performance under FGSM adversarial attacks across epsilon values from 0.01 to 0.8. Found that synaptic filteri…

    PythonPyTorchTransformersLLaMA 3FGSM
  • Machine Learning

    Multi-Task NLP via Task Vector Composition

    Built a parameter-efficient multi-task learning system using task vector composition, reaching 87% accuracy on sentiment analysis and 74% F1 on named-entity recognition. Held the model to 110M paramet…

    PythonPyTorchTransformersBERTHuggingFace

Get in touch

Let's connect

Reach out via email or on social media — I'd love to hear from you.