The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Back to feedPost
The challenge
Complex property-intelligence requests rarely fit into a single prompt. They can involve market research, asset analysis, knowledge retrieval, recommendation generation, and structured reporting.
A general-purpose chatbot was not sufficiently controllable for these workflows. It could lose context, select the wrong tool, produce inconsistent recommendations, or fail without providing enough information for engineers to diagnose the problem.
My role
As the AI Engineer and Agentic Systems Architect, I designed and productionized the system end to end—from workflow decomposition and agent routing to retrieval, validation, deployment, and evaluation.
What I built
Specialized agents with clearly scoped responsibilities and tool contracts
Dynamic routing, task planning, working memory, and agent handoffs
Hybrid retrieval and reranking for evidence-grounded responses
Structured outputs with schema validation and quality checks
Guardrails, fallback paths, retries, and human-escalation points
Python and FastAPI services deployed through Docker and Kubernetes
Centralized logging, metrics, tracing, and failure monitoring
Evaluation workflows tied to representative business tasks
Result
The platform achieved more than 85% end-to-end task completion across representative business workflows.
More importantly, it transformed one-off agent logic into reusable production patterns. New workflows could reuse the same routing, retrieval, validation, observability, and recovery infrastructure instead of rebuilding the agent stack from scratch.
This case study includes only non-confidential architectural details.
DETAILS
Role: AI Engineer / Agentic Systems Architect
Client or organization: PropTy Global
Duration: August 2024–September 2025
Industry: Real Estate, SaaS
Roles: AI Developer, Machine Learning Engineer, AI Consultant
Tools and skills: Python, FastAPI, RAG, LangChain, Docker, Kubernetes
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Trending
retrofuturism
Share your favorite designs from the past exploring the distant future. Or your own take on the aesthetic 🚀
halftone
Dots. Texture. Grain. Let's see your designs using the halftone technique.
Midjourney
Creatives are showing off what they're making with Midjourney.
colorgrade
Video editors sound off: How do you enhance the mood of your video in color grading?
Related posts
A government just hired us to build AI that screens foreign investments for national security risks.
Here's what happened.
Ukraine's government launched a program funded by a UK grant to modernize how it evaluates foreign investment risk. They needed an AI engine that could estimate financial risk across multiple variables, fast and at scale.
We're excited to be part of that transition.
For us this is a different kind of project. We've built 40+ AI and automation systems for founders and startups across the US, UK, and Scandinavia.
But a government choosing you to build something this sensitive says one thing clearly: the work you did before was trusted enough to bet national policy on.
Here's what I'm noticing.
Governments are no longer "exploring AI." They are buying. They are shipping. They are putting it into decision-making pipelines that actually matter.
That's the shift.
If you've been building real AI systems for real businesses, your next client might not be a startup.
It might be a country.
Agentic Financial Document Intelligence & Transaction Reconciliation Engine
• Role: Lead AI & Systems Architect
• Deliverables: Multi-Modal PDF Parser, Hybrid Agent Orchestrator, Deterministic Math Engine, Audit Trail Logger
• Timeline: 3 Weeks
OVERVIEW
Financial institutions and back-office teams spend thousands of manual hours reconciling unstructured documents (loan contracts, invoices, compliance bylaws) with relational transaction ledgers. Standard LLM wrappers fail catastrophically in this domain due to mathematical hallucinations, version conflicts, and a lack of explainable audit trails.
I engineered an agentic document intelligence system that combines multi-modal PDF extraction with deterministic validation tools to automate complex cross-referencing with 100% computational precision.
THE PROBLEM & DIAGNOSIS
LLM Math Hallucinations: Generative models cannot be trusted to perform exact financial calculations, interest compounding, or tax audits.
Unstructured vs. Tabular Asymmetry: Document facts are scattered across messy layouts, whereas transaction logs live in strict relational schemas.
Document Version Discrepancies: Clauses change across revisions; systems must resolve conflicting policies before executing reconciliation.
KEY ARCHITECTURAL DECISIONS
• Decision 01: Separation of Reasoning and Execution
The LLM acts strictly as a reasoning and extraction layer, identifying entities and formulas. All financial calculations and reconciliations are handed off to deterministic Python execution sandboxes.
• Decision 02: Cross-Source Entity Disambiguation
Built a cross-referencing engine that maps unstructured text snippets directly to SQL transaction IDs using hybrid vector-keyword scoring.
• Decision 03: Fully Auditable Chain-of-Thought
Every validation output includes a step-by-step citation trail pointing to the exact page, bounding box, and ledger entry.
MEASURABLE RESULTS & IMPACT
• 0% computational hallucination rate achieved via deterministic code execution.
• 90%+ precision in entity extraction across multi-page, complex financial agreements.
• Production-grade audit trails satisfying compliance requirements.
TECH STACK
Python 3.11, FastAPI, Pydantic, LangChain, PyMuPDF, Pandas, PostgreSQL / pgvector, Docker.
Day 4 of learning and creating with AI.
Today, I created a UGC-style product campaign for AURA PRO, a fictional wireless earbuds brand.
The goal wasn't just to make the product look good. I wanted the video to feel natural, relatable, and believable — like a real person sharing why they chose the product.
From the product concept to visuals → AI model → storytelling → video, everything was built with AI.
Still experimenting. Still learning. Still getting better.
Day 4/30.
What do you think?
AI Engineer Adobe IllustratorChatGPTGoogle FlowaivideocreationugccreatoraiproductphotographyAI Video ProductionUGC Creation
The focus on believability is exactly right for UGC. A polished render is easy to notice, but natural pacing and human context are what make a product story feel trustworthy.