All Studio
SaaS fitness CRM that flags at-risk clients in real time — no more weekly manual reviews.
ReactSpring BootPostgreSQLJWTTailwind CSSVite
builder
Full-stack systems from idea to production
<engineer />
AI pipelines, ML systems, and agentic tooling
Koshik Kumaravel
AI Engineering intern at Fuel Cycle and ML researcher at UCI, building full-stack systems that put AI to work.
Portfolio
Selected Work
Movie Recommender
Hybrid ML engine combining collaborative filtering and content-based models, benchmarked across five algorithms.
PythonPandasNumPyScikit-learnSciPyJupyter Notebook
Background
About Me
TODO: Write a 2–3 sentence bio for the About section.
Currently studying B.S. Artificial Intelligence at University of California, San Diego (GPA: 3.92, expected June 2028).
Languages
PythonJavaC++JavaScriptTypeScriptSQLHTML/CSSPHP
Frameworks & Libraries
ReactNode.jsExpressSpring Bootscikit-learnPandasNumPyMatplotlib
Tools & Practices
GitDockerPostmanLinuxBashREST APIsCI/CDAgile
AI & ML
LLM integrationMCPAgent2AgentPrompt engineeringRecommender systems
Experience
AI Engineering Intern
Fuel Cycle · Los Angeles, CA
- Built an agentic QA platform that auto-generates tests from Jira tickets using MCP — cutting QA cycle time by 60%.
- Produced 50+ adversarial, regression, and security tests per API ticket, catching critical auth bugs before release.
- Pitched the architecture to executive leadership; it was adopted as the team standard.
Machine Learning Research Assistant
UCI ProperData Research Team · Irvine, CA
- Built a personality inference pipeline that maps datasets to Big Five trait profiles using a locally-hosted LLM.
- Designed prompts grounded in published annotation protocols, reducing output token size by 51%.
- Created an interactive visualization tool with a live OCEAN radar chart and real-time confidence scoring.
Software Engineering Intern
Priyum Products · Irvine, CA
- Built a Flask pipeline that auto-detects and redacts weight data from shipping label PDFs — no more manual cropping.
- Integrated Shippo and Shopify APIs to collapse a 4-step label workflow into a single CLI command.
- Dockerized the service and deployed on Render; handles multi-page PDFs under 400MB with zero on-disk storage.
Say hello
Get In Touch
Open to internship and new grad opportunities in software engineering and AI. Feel free to reach out.