I turn ideas into intelligent software
AI engineer and full-stack developer. I build multi-agent systems, LLM apps with schema-enforced output and the benchmarks that test them β then ship them as real products with Next.js, TypeScript and FastAPI.
At a Glance
π§ͺ AI Eval
@ Handshake AI
π 4
Live AI Apps
β‘ Full-Stack
Next.js + FastAPI
π°π· EN+KR
Bilingual Developer
About
Available ImmediatelySweetyCodes
AI Engineer & Full-Stack Developer
AI engineer and full-stack developer. I build AI systems that hold up in the real world β multi-agent pipelines, LLM apps with schema-enforced output β and the full-stack products that put them in front of real users.
As an AI Evaluation Engineer via Handshake AI, I design agentic benchmark tasks that evaluate AI coding agents. On my own, I've shipped InterviewLens, a 5-agent interview simulator, and K-Trust, a compliance portal that returns typed JSON from Gemini β both built end to end with Next.js, TypeScript, Python and FastAPI. Reliability is what I care about most: retries and fallbacks, confidence-gated writes, and evals that separate real model failures from bugs.
Education
B.Sc Computer Science Β· 8.41 CGPA
Location
Open to Remote Globally
Status
Available Immediately
What I Build
Multi-Agent Systems
5-agent pipelines with memory, retries & fallbacks
AI Evaluation
Agentic benchmarks with Dockerized verifiers
Structured LLM Output
Schema-enforced JSON via Pydantic + Gemini
Agent Reliability
Confidence-gated writes and graceful failure
Full-Stack Web Apps
Next.js, React & TypeScript β from UI to deploy
APIs & Backends
FastAPI + Pydantic services, Firebase & SQL data
My Journey
- π2023
Started B.Sc Computer Science
Where it all began
- β‘2024
Discovered Frontend Development
Built first web projects and fell in love with UI
- π2025
Deep Dived into Python & AI
100+ hours of self-learning + KMOOC Data Science & AI certificate
- πJan 2026
3 Hackathon Podiums
1st Code Relay Β· 1st Solution Sprint Β· 2nd Bug Buster
- π2026
Graduated & Shipping AI Apps
8.41 CGPA β InterviewLens, K-Trust, Draping, CarbonIQ
- π€Jul 2026
AI Evaluation Engineer
Building benchmarks that test AI coding agents β via Handshake AI
Featured Builds
Full-stack AI apps β live, on GitHub, and built end to end.
5-agent AI interview simulator with memory that compounds across sessions.
Audits Korean merchants and returns trust scores as schema-enforced JSON.
AI personal colour analysis from a selfie, plus virtual outfit try-on.
Log your footprint in plain language; low-confidence AI reads ask before saving.
Highlights
Live AI apps
InterviewLens, K-Trust, Draping and CarbonIQ β all live and on GitHub
Multi-agent pipelines
Five agents with memory, retries and fallbacks
Hackathon podiums
Two 1st places and a 2nd, in one month
Self-taught in AI
100+ hours, plus the KMOOC Data Science & AI certificate
Tech Stack
AI & LLMs
Agents & Evaluation
Languages
Frontend
Backend & Data
Python Libraries
Dev & Deploy
AI Dev Tools
Currently Learning
What I Can Build For You
AI Agents & Multi-Agent Systems
Agent pipelines with tool use, memory, retries and fallbacks β built on Gemini and OpenAI
LLM Apps with Structured Output
AI features that return typed, schema-validated JSON your frontend can trust
AI Evaluation & Benchmarks
Benchmarks and automated verifiers that test how well LLMs and AI agents really perform
Full-Stack Web Apps
Production Next.js, React and TypeScript apps β from UI to deployment
APIs & Python Backends
FastAPI services with Pydantic models, plus Firebase and SQL data layers
Bilingual EN/KO Products
Fully localized English and Korean experiences, end to end
Work
EXPERIENCE
AI Evaluation Engineer (Contract)
via Handshake AI
- Designed and built agentic, terminal-based benchmark tasks to evaluate AI coding agents, including Dockerized environments, reference solutions, and automated verifiers
- Calibrated task difficulty empirically by measuring real oracle output variance to set verification tolerances, ensuring tasks reliably distinguish genuine agent failure from task/infra bugs
- Navigated a multi-stage automated review pipeline (static checks, rubric review, similarity detection, pass@2 and pass@5 agent trial gating) to deliver production-ready evaluation tasks end-to-end
Business Website
Ready-to-launch templates for small businesses β lead generation, e-commerce and SaaS. Designed, built and deployed end to end.
CarbonShine Detailing
Lead GenPremium dark mode car detailing studio with pricing cards and working lead capture form.
Project Details
Full-featured lead generation site for a detailing business. Includes dynamic pricing, testimonial sliders, and an automated contact form.
Paws & Play
E-CommerceSmart pet toy product page with image gallery and Stripe test payment integration.
Project Details
E-commerce concept for smart pet technology. Integrated with Stripe for seamless payment processing and a responsive product gallery.
TaskFlow AI
SaaSAI-powered freelance automation SaaS landing page with FAQ accordion and waitlist form.
Project Details
SaaS platform designed to automate freelance workflows using AI. Features a waitlist system and interactive FAQ components.
CarbonShine Detailing
Lead GenPremium dark mode car detailing studio with pricing cards and working lead capture form.
Project Details
Full-featured lead generation site for a detailing business. Includes dynamic pricing, testimonial sliders, and an automated contact form.
ACHIEVEMENTS
1st Place Solution Sprint
Intercollege Hackathon β’ Jan 2026
1st Place Code Relay
Intercollege Hackathon β’ Jan 2026
2nd Place Bug Buster
Intercollege Hackathon β’ Jan 2026
Keynote Speaker Blockchain Seminar
Khallikote University β’ Year 1
CERTIFICATIONS
KMOOC Data Science and AI
Oct 2025
Full Stack Web Development
Self-Learned β FreeCodeCamp + CS50W
Git and GitHub Version Control
Self-Learned β Practiced Daily
EDUCATION
B.Sc Computer Science
Khallikote University
8.41 CGPA
Senior Secondary XII Science CBSE
Kendriya Vidyalaya
88%
Secondary X CBSE
Kendriya Vidyalaya
91.4%
Why Work With Me?
I evaluate AI for a living
I build benchmarks that tell real AI agent failures apart from bugs.
Reliability is built in
Retries, fallbacks and confidence-gated writes in the AI apps I ship.
AI and full-stack in one person
From agent pipeline to Next.js UI to FastAPI backend β no hand-offs needed.
Everything is live
4 AI apps deployed and on GitHub β you can try every one of them right now.
Proven under pressure
Three hackathon podiums in one month β two 1st places and a 2nd.
Bilingual EN / KR
Fluent English, conversational Korean, and preparing for TOPIK.
Let's Build What's Next
Open to AI engineer and full-stack roles, internships and freelance AI projects β remote
sweety22313@gmail.com