You will get a production AI agent built with Claude, MCP, and real tests
Project details
Most AI agents demo beautifully and break in production — they hallucinate, loop, call the wrong tool, leak data, or fall over on the first real edge case.
I build agents that hold up. I design and ship production AI agents with Claude, the Agent SDK and MCP: tool-using and agentic workflows, RAG and knowledge agents, support and ops automation, and multi-agent systems — on any stack.
Every build ships with a real eval/test suite (not a happy-path demo), guardrails against prompt injection and tool abuse, proper error handling and retries, memory/state, and clean documented code. Higher tiers add deployment and monitoring so you can see what your agent does in production.
Why me: I also audit AI agents for reliability and security (1st-place Solana security auditor, CVSS 7.5; 125+ vulnerabilities documented), and I ship agents backed by real test suites — one production agent has 287 automated tests. Builder and breaker in one, so the failure modes get designed out from the start.
Specialty in on-chain / Web3 agents. (I don't build trading, market-making or copy-trading bots.)
Fully async, written-first. Message your use case for a quick fit check before you order.
I build agents that hold up. I design and ship production AI agents with Claude, the Agent SDK and MCP: tool-using and agentic workflows, RAG and knowledge agents, support and ops automation, and multi-agent systems — on any stack.
Every build ships with a real eval/test suite (not a happy-path demo), guardrails against prompt injection and tool abuse, proper error handling and retries, memory/state, and clean documented code. Higher tiers add deployment and monitoring so you can see what your agent does in production.
Why me: I also audit AI agents for reliability and security (1st-place Solana security auditor, CVSS 7.5; 125+ vulnerabilities documented), and I ship agents backed by real test suites — one production agent has 287 automated tests. Builder and breaker in one, so the failure modes get designed out from the start.
Specialty in on-chain / Web3 agents. (I don't build trading, market-making or copy-trading bots.)
Fully async, written-first. Message your use case for a quick fit check before you order.
AI Algorithms
Large Language Model, Multimodal Large Language ModelAI Applications
AI Chatbot, Conversational AIAI Development Language
PythonAI Models
ChatGPT, GPT-4, LLaMAWhat's included
| Service Tiers |
Starter
$499
|
Standard
$1,200
|
Advanced
$2,600
|
|---|---|---|---|
| Delivery Time | 5 days | 12 days | 21 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 |
Frequently asked questions
About Rheza
AI Agent Engineer on Solana | Audit-Proven | RECTOR LABS
Bekasi, Indonesia - 12:00 pm local time
I don't build chatbots that get stuck. I build agentic systems that survive mainnet: LLMs that orchestrate real on-chain actions, agent privacy infrastructure, multi-chain stealth tooling. Every line of agent code I ship is audited like a smart contract — because that's what agents are: programs with money and autonomy.
What I build:
- Production agentic apps — LLM tool calling, multi-step reasoning, auto-recovery, preflight simulation
- Agent infrastructure on Solana — REST APIs + skill files for AI agents to consume on-chain primitives
- Solana programs — Rust + Anchor, audit-grade architecture, on-chain logic that survives mainnet
- Agent privacy — stealth addresses, viewing keys, Pedersen commitments across 17 chains
- Smart contract security — vulnerability hunting, audit reports, mitigations (13 findings across 14 audited Solana repos)
Proof of work:
- Sipher — Privacy-as-a-Skill for Multi-Chain Agents (66 endpoints, 17 chains, on-chain Anchor program, 1,402 tests, live in production)
- Kami — AI Co-Pilot for Kamino DeFi (plain English → signed mainnet tx, LLM auto-recovery, live on Solana mainnet)
- 11 hackathon & bounty wins — $36,050+ across the Solana ecosystem
- $10K Solana Foundation grant + $6K audit subsidy for SIP Protocol
- 1st of 116 in a Solana Security Audit — 13 findings across 14 repos, incl. a framework-level Anchor CPI bug (CVSS 7.5, fixed upstream)
How I work:
- Production-first — every line written as if it ships tonight, audited tomorrow
- Fully async — written updates + Loom walkthroughs by default. I'm in Jakarta (UTC+7); async keeps us both unblocked across time zones. Saves you status meetings, gives me deep focus.
- Honest comms — clear timelines, no scope creep, no ghosting
- Security-conscious — agents move money, so the bar is high
- Pragmatic — I ask questions before coding, pick the right tool, and say "no" when a feature doesn't serve the goal
Stack: Solana, Rust, Anchor, TypeScript, Next.js, Vercel AI SDK, MCP, Noir (ZK), Python, Docker, PostgreSQL
Let's ship a production agent — for your DeFi protocol, your trading desk, your DAO treasury, or whatever else needs to be smart, secure, and on-chain.
Steps for completing your project
After purchasing the project, send requirements so Rheza can start the project.
Delivery time starts when Rheza receives requirements from you.
Rheza works on your project following the steps below.
Revisions may occur after the delivery date.
Scope & design
Lock the agent's goal, tools, control flow, guardrails, and success criteria.
Build the agent
Reasoning loop, tool/MCP integrations, memory/state, and error handling with retries.