You will get your broken AI agent fixed: audit, guardrails, evals
Project details
Your AI agent worked in the demo. Now it hallucinates, ignores instructions, or takes wrong actions - and nobody can explain why. I diagnose and fix production AI systems; as lead engineer on a consumer AI product, making agents reliable is my day job. Send me your agent (code, prompts, a way to reproduce failures). You get a plain-English root-cause report - usually retrieval quality, prompt architecture, missing guardrails, no eval coverage, or the wrong model - plus a prioritized fix list. Higher tiers: I implement the fixes and leave an eval suite that catches regressions before users do.
AI Development Type
Deep Learning, Knowledge Representation, Model Tuning, Recommendation System, Software MaintenanceAI Development Language
PythonWhat's included
| Service Tiers |
Starter
$490
|
Standard
$1,250
|
Advanced
$2,450
|
|---|---|---|---|
| Delivery Time | 5 days | 10 days | 14 days |
Number of Revisions | 1 | 2 | 2 |
AI Model Integration | |||
Detailed Code Comments | |||
Knowledge Graph | |||
Model Documentation | |||
Ontology | |||
Source Code | |||
Taxonomy |
1 review
(1)
(0)
(0)
(0)
(0)
This project doesn't have any reviews.
DD
Dawson D.
Dec 18, 2022
Backend for Webapp
Anas worked fast and communicated frequently and made sure I was updated. His code is clean and well documented, it was a pleasure to work with him.
About Anas
Agentic AI Engineer | AI Agents, Automation, RAG & LLM Apps
Islamabad, Pakistan - 11:20 am local time
Most AI agents break the moment they meet real users, real data, and real edge cases. I build the unglamorous parts that make them survive: evals, guardrails, fallbacks, confirmation gates, and clean integration into your existing stack - so you get a system you can trust, not a black box you're afraid to touch.
WHAT I BUILD
• Autonomous AI agents & multi-agent systems - tool/function calling, orchestration, memory
• RAG systems & AI chatbots grounded in your documents and live app data, with cited answers
• LLM features inside existing products (OpenAI, Anthropic Claude, Gemini, open-source)
• Agent guardrails, evals & observability - so you can see and trust what's running
• The backend behind it all: Node.js, NestJS, Go, PostgreSQL, Redis, Kafka, GCP
PROOF
I currently lead engineering for a consumer family-safety AI product: I architected the context layer that grounds the assistant in live app state, designed the guardrails that let agents act safely on a parent's behalf, and scaled to 30+ audited agent actions in production. Earlier: cut an API's latency 70% (1.2s to ~350ms) for a platform serving 1,000+ users, and shipped checkout features that helped a fintech startup close a $10M round. On Upwork: $50K+ earned across long-term engagements, 5.0 rating, 1,900+ hours.
HOW I WORK
I start with the workflow, not the tech. If an agent pays for itself, I'll tell you how I'd build it - architecture, effort, and monthly running costs. If it doesn't, I'll tell you that too, before you spend money. Every build ships with evals, monitoring, documentation, and a clean handover. I also write about agent reliability, observability, and memory vs. context on my engineering blog.
Have a manual, repetitive, expensive workflow that should run itself - or an AI agent that isn't behaving in production? Message me with what's eating your team's time, and I'll tell you exactly how I'd approach it.
Steps for completing your project
After purchasing the project, send requirements so Anas can start the project.
Delivery time starts when Anas receives requirements from you.
Anas works on your project following the steps below.
Revisions may occur after the delivery date.
Requirment Gathering
Meetings with various stakeholders to define the product.
Execution
Work in progress, with timely updates