You will get custom AI agents & multi-agent systems, production-ready and deployed
Project details
AI agents that actually do work — and ask before they act. Each uses only the tools you allow and pauses before anything risky.
WHAT THAT LOOKS LIKE
A planner splits your request; sub-agents work in parallel — searching your docs, querying your DB, calling the web — then pause at an approval gate. Nothing is sent or changed until you click Approve. Every step logged, every answer cited.
WHAT I BUILD
• Agents for real tasks: research, monitoring, reports, data entry, follow-ups
• They use tools — web, your DB, your APIs — and can draft & send messages, schedule meetings, place calls, control apps and browsers. Most things a person does on a computer, an agent can automate.
• Multi-agent systems coordinated by a planner
• Deployed on a schedule, webhook, your API, a site widget, or Slack
• Models: GPT, Claude, Gemini, or open-weight Llama/Qwen on your servers
WHY ME
Senior AI engineer, 7 years, ex-CTO. Production AI on my own GPU infra — approval gates, guardrails, logging, real testing. I test against your tasks and hand you a behavior report.
Not sure how to apply this? Message me — I'll map your use case and quote it, free. Fixed price, escrow milestones, async (GMT+9).
WHAT THAT LOOKS LIKE
A planner splits your request; sub-agents work in parallel — searching your docs, querying your DB, calling the web — then pause at an approval gate. Nothing is sent or changed until you click Approve. Every step logged, every answer cited.
WHAT I BUILD
• Agents for real tasks: research, monitoring, reports, data entry, follow-ups
• They use tools — web, your DB, your APIs — and can draft & send messages, schedule meetings, place calls, control apps and browsers. Most things a person does on a computer, an agent can automate.
• Multi-agent systems coordinated by a planner
• Deployed on a schedule, webhook, your API, a site widget, or Slack
• Models: GPT, Claude, Gemini, or open-weight Llama/Qwen on your servers
WHY ME
Senior AI engineer, 7 years, ex-CTO. Production AI on my own GPU infra — approval gates, guardrails, logging, real testing. I test against your tasks and hand you a behavior report.
Not sure how to apply this? Message me — I'll map your use case and quote it, free. Fixed price, escrow milestones, async (GMT+9).
AI Algorithms
Large Language Model, Multimodal Large Language Model, Transformer ModelAI Applications
Conversational AI, Natural Language Generation, Natural Language UnderstandingAI Development Language
PythonAI Tools
Azure OpenAI, Hugging Face, NVIDIA AI Platform, PyTorchAI Models
ChatGPT, GPT-4, LLaMAWhat's included
| Service Tiers |
Starter
$150
|
Standard
$700
|
Advanced
$1,500
|
|---|---|---|---|
| Delivery Time | 5 days | 12 days | 25 days |
Number of Revisions | 1 | 2 | 3 |
AI Model Integration | |||
Batch Normalization | - | - | - |
Database Integration | - | ||
Detailed Code Comments | - | - | - |
Image Upscaling | - | - | - |
MLOps | - | - | |
Model Deployment | |||
Model Documentation | - | - | - |
Model Monitoring | - | - | |
Model Testing & Optimization | - | ||
Model Tuning | - | - | - |
Natural Language Processing | |||
NLP Tokenization | - | - | - |
Pre-Training | - | - | - |
Prompt Engineering | |||
Setup File | - | - | - |
Source Code |
Optional add-ons
You can add these on the next page.
Fast Delivery
+$60 - $500
Additional Revision
+$80
Additional specialized subagent
(+ 4 Days)
+$500
RAG / knowledge subagent
(+ 5 Days)
+$600
On-prem / self-hosted LLM (vLLM)
(+ 7 Days)
+$1,200Frequently asked questions
About Chris
Production AI Engineer | RAG, AI Agents & Integration
Seoul, South Korea - 12:50 am local time
For founders and product teams, I build and repair RAG, AI-agent, and OpenAI/Claude systems with measured quality, cost controls, observability, and human approval for sensitive actions.
GOOD FIT IF YOU NEED
• A RAG flow with source citations, safe fallback, and an evaluation set
• A LangGraph/PydanticAI agent with tool calling and approval gates
• An OpenAI or Claude API issue fixed in Python/FastAPI
• One AI feature integrated into an existing Next.js product
PROOF
• 7+ years shipping applied AI; former founding CTO of a Seoul AI startup
• 60+ paid AI projects delivered through major Korean freelance marketplaces
• Cut a review-and-outreach workflow from ~60 minutes to ~5
• Improved multimodal retrieval speed by 98.4% and reduced GPU/database cost by 80%
• Shipped review-analysis models inside CJ OnStyle
• 6 registered patents and a peer-reviewed ML paper
HOW I WORK
We agree on the failure mode, scope, and proof of completion before I build. You receive reviewable increments, tests, clear tradeoffs, concise status updates, and handoff documentation.
Before hiring or ordering a Project Catalog package, send me an Upwork message with your current stack, what is failing or missing, and the outcome you need. I’ll confirm fit and recommend the smallest practical next step.
Steps for completing your project
After purchasing the project, send requirements so Chris can start the project.
Delivery time starts when Chris receives requirements from you.
Chris works on your project following the steps below.
Revisions may occur after the delivery date.
Discovery & scoping
I review your goal and tools, recommend the model setup, and send a fixed-price offer with clear milestones.
Agent design
Define each agent's job, its allowed tools, and its limits — including where it must stop and ask a human.