Agentic AI
Agent skills, LLM evaluation and model validation.
Hi, I'm
Jeevika Kiran
AI & Software Engineer
MSCS at University of Southern California
I build intelligent developer tools, scalable AI systems and multimodal applications, bringing together industry engineering and collaborative research to create technology that matters.
About me
I’m drawn to the things we haven’t figured out yet.
I'm Jeevika, a graduate student pursuing an M.S. in Computer Science at USC. I'm particularly interested in AI research and development—and in turning emerging ideas into systems that are genuinely useful. Through research collaborations during my undergraduate studies and at USC, I've explored new technologies and how they can address real-world problems.
Most recently, in Summer 2026, I interned as an AI Foundations Engineer at the IBM T. J. Watson Research Center, working on Agent Skills and evaluation methods for AI agents. Before graduate school, I built infrastructure and backend services for production AI systems at IBM Software Labs.
How I work
I'm happiest when the answer isn't obvious: understanding a new idea deeply, experimenting with intention, then turning what works into reliable, practical technology.
AI Systems
RAG, inference workflows and production Python services.
Multimodal ML
Vision, audio and learning systems for robotics.
At Scale
Data infrastructure processing more than 1M daily events.
Experience
Where I've
Built Impact
A timeline of my professional journey so far.
Download resumeMay 2026 — Aug 2026
AI Foundations Software Engineer Intern
IBM Research · T. J. Watson Research Center
Yorktown Heights, NY
Built AI developer tools and reusable Agent Skills for the IBM Granite Agent Laboratory, automating benchmark evaluation, failure analysis and model validation. Developed full-stack workflows that generated interactive HTML, Plotly and Streamlit dashboards through IBM Bob, alongside CI/CD pipelines for compatibility testing across Hugging Face, Transformers and ONNX.
PythonFastAPIOllamaRabbitMQMongoDBStreamlitPlotlyHugging FaceTransformersONNXCI/CDIBM Granite
Key impact
View work- Reduced manual feedback analysis by 70% with interactive dashboards
- Improved Granite compatibility-testing accuracy by 50%
- Validated models across Hugging Face, Transformers and ONNX
Projects
My Projects
A selection of products and experiments shaped by curiosity, care,
and a drive to solve meaningful problems.
AI & ML
Deepfake Detection
A multimodal deepfake detection system that fuses visual artifacts, audio authenticity and lip-sync alignment into one per-video probability, reaching 0.9993 ROC-AUC and 98.67% validation accuracy on FakeAVCeleb.
PyTorch · EfficientNet-B0 · Wav2Vec2 · SyncNet · MTCNN · OpenCV
AI & ML
Life Lens
An AI-guided VR therapy platform that lets therapists design multi-stage sessions and delivers them on Meta Quest through immersive 360° environments, Gemini Live voice guidance and intelligent scene transitions.
Gemini Live · WebXR · Three.js · React · FastAPI · Meta Quest
AI & ML
Foresight
Foresight is an open-source accessibility accelerator for visually impaired users, integrating ShareGPT4V for visual analysis, Detectron2 for object grounding, and Gemma 7B for improved accuracy. With speech-to-text and text-to-speech interaction, it reduces hallucinations and identifies 12% more useful information than baseline models, delivering reliable, context-aware assistance.
Generative AI · Detectron2 · ShareGPT4V · Gemma 7B · ReactJS · Django · SQLite · Kotlin
Systems
AuthEZ
AuthEZ is a passwordless identity provider SDK that combines SSO, facial verification with liveness checks, voice embeddings and RSA cryptographic signatures for secure, seamless access.
Identity Provider · Facenet · Resemblyzer · WebAuthn API · ReactJS · JavaScript · Tailwind CSS
Web Development
Memoirs
A mental-health surveillance and assessment platform for teens and students, with personalized music recommendations and AI-powered assistance.
React · TypeScript · Next.js
Web Development
Student Buzz
A collaborative platform for students to showcase their skills and projects, participate in club activities and connect with peers across colleges.
Django · ReactJS · SQLite
AI & ML
Car Theft Detection System
An automated vehicle-theft detection system using moving-object detection and barcode scanning to identify suspicious movement and alert parking personnel.
Python · OpenCV · Computer Vision · Digital Image Processing
Systems
Valunique
A bias-aware recruitment platform with faceless hiring and role-based candidate and admin portals.
ReactJS · NodeJS · Microsoft SQL
Achievements
Milestones That Define My Journey
A collection of recognitions, publications and certifications that reflect my passion for learning, research and impact.
Publications
3 publications · 2 peer-reviewed
Awards & Recognitions
2026
IBM Internship Certificate
2024
Best Capstone Project
2024
Research Recognition
2023
Top 7 / 700
Unisys Innovation Program. Global finalist with AuthEZ, selected from more than 700 participating teams.
Certifications
2026
AI Engineer Agentic Track
Udemy
Complete Agent & MCP course covering agentic systems and reusable AI tooling.
2024
Enterprise Design Thinking
IBM
Applying enterprise design-thinking principles to practical product challenges.
2023
Convolutional Neural Networks
Coursera
Deep-learning concepts and applications for visual recognition systems.
2023
Advanced Learning Algorithms
Coursera
Advanced machine-learning algorithms, evaluation and implementation practices.
2022
Supervised Machine Learning
Coursera
Regression and classification techniques for supervised-learning problems.
2022
Foundational Data, ML & AI
Google Cloud
Foundational data, machine-learning and AI workflows on Google Cloud.
2021
Python Data Structures
Coursera
Core Python data structures and their use in practical problem solving.
Contact
Let's create something meaningful.
I'm always interested in thoughtful products, ambitious ideas and new opportunities. Tell me what you're working on—or simply say hello.
Start a conversation
Send me a message
Usually replies within two business days.