You will get I will build a custom AI agent to automate your workflows
Project details
You will get a production-ready AI automation system that eliminates repetitive workflows and scales your operations. I specialize in building full-stack AI agents and workflow orchestration systems using LLMs, API integrations, and modern automation tools. With expertise in prompt engineering, RAG systems, and agentic AI frameworks, I deliver solutions that integrate seamlessly with your existing stack. 100% test coverage, clean documentation, and knowledge transfer included.
AI Algorithms
Large Language Model, Recurrent Neural Network, Transformer ModelAI Applications
AI Chatbot, Conversational AI, Natural Language GenerationAI Development Language
PythonAI Tools
Azure OpenAI, Hugging FaceAI Models
OpenAI CodexWhat's included
| Service Tiers |
Starter
$300
|
Standard
$800
|
Advanced
$2,000
|
|---|---|---|---|
| Delivery Time | 7 days | 14 days | 21 days |
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 | - | - | - |
About Jody
AI Automation Engineer | Claude, LangChain, n8n, RAG & AI Agents
Hong Kong, Hong Kong - 8:09 pm local time
Most agent projects fail the same way: it works in testing, then quietly degrades and nobody can say why. I build the measurement in from the start.
A production agent I built and operate:
• LangChain/LangGraph agent across four tool capabilities — web search, sandboxed code execution in Docker, retrieval, third-party APIs
• Provider-agnostic: one config string switches between Anthropic, OpenAI, Google, NVIDIA and MiniMax
• Full trace on every call — input, output, latency, token usage, tools invoked, call provenance
• Evaluation harness: 21-item golden dataset, rule graders plus an LLM judge, batch re-runs computing pass@1 and pass^k, version-to-version comparison
• Measured, not estimated: pass@1 96.2%, pass^5 81.0% (held-out 96.7% / 83.3%), judge validity at Cohen's kappa = 1.000
• Guardrails on the live path, not only offline — flagged into the trace for human review
• Hosted behind FastAPI with Bearer-key auth: keys hashed at rest, revocable, per-key rate limiting
What I build:
→ AI agents and multi-agent orchestration (Claude, OpenAI, LangChain/LangGraph, MCP)
→ RAG pipelines — document parsing, semantic search, citation-backed answers
→ Business process automation with n8n, APIs and webhooks — including validation, retries, error handling and logging
→ Unattended data pipelines — daily multi-source ingestion, transformation and validation
→ Full-stack AI apps (Python/FastAPI + TypeScript/React/Next.js)
Also shipped: a Next.js 15 + LangGraph planning app with a multi-tier Redis cache that cut API response times by over 50%, and a Python system with 1,015 passing tests at 80% coverage.
One thing I care about. I once had a system of mine run away with token consumption. What came out of it is how I now build — pinned model configuration, privileged options removed from any automatic fallback path, manual-only escalation, and a daily drift audit that raises a CRITICAL alert on unauthorised change. If you are putting an agent anywhere near your live business systems, this is the part that matters.
Before going independent I spent 14 years delivering IT and AI projects, most recently as Senior Project Manager at SenseTime, where I delivered AI vision projects for MTR, the Airport Authority and HK Land. I scope properly, I write things down, and I tell you early when something will not work rather than late.
Tech: Python, TypeScript, Node.js, FastAPI, LangChain, LangGraph, Claude API, OpenAI API, n8n, Docker, Kubernetes, Pinecone, Chroma, FAISS, Redis, SQLite, CDP, CI/CD
Working GMT+8, available 30+ hrs/week, fluent English. Tell me what you need built or automated — I will tell you honestly whether I am the right person for it.
Steps for completing your project
After purchasing the project, send requirements so Jody can start the project.
Delivery time starts when Jody receives requirements from you.
Jody works on your project following the steps below.
Revisions may occur after the delivery date.
Define scope, approach, and deliverables. Create architecture plan.
Requirements Analysis & Planning