Elroy Reinhard
Full-stack web apps with AI inside, shipped for a large Indonesian state-owned bank. Hour-long workflows turned into minutes, with humans in the loop.
AI Systems Shipped in Production
End-to-end architectures operating in regulated banking environments. Anonymized to preserve organizational confidentiality.
Contract Drafting Assistant
Turning hour-long procurement contract preparation into a 2-minute verified workflow, used by two divisions.
Knowledge Assistants (RAG Chatbots)
Modernized and scaled operational knowledge retrieval for operations and branch staff, cutting answer latency and refusing 100% of out-of-scope questions in evaluation.
Onboarding Form Verifier (OCR)
Pairing vision OCR with alias matching to verify corporate onboarding forms and surface only the fields that need review.
What I Can Build for You
I build web applications with AI inside, the kind that remove hours of manual work.
Document Automation
From multi-page source documents to production-ready drafts, with human-in-the-loop verification.
Knowledge Assistants
Grounded chatbots over enterprise internal documents, with page citations, out-of-scope refusal, and a safe knowledge update lifecycle.
Document Verification (OCR)
Comparing physical paperwork against digital records, surfacing only field discrepancies.
Web Apps & Internal Tools
Full-stack web applications and internal tools, from one module to a multi-module platform with single sign-on, role-based access, and audit logs.
How we'd work together
- 01
Scope first
we map the manual workflow and agree on one measurable result before anything is built.
- 02
Fixed-price phases
work is split into phases, each with a clear deliverable and acceptance criteria.
- 03
Built with checks
prototypes run on your real documents, results are evaluated, and people verify what the AI produces.
- 04
Clean handover
you receive the source code, documentation, and a working deployment.
- 05
Optional maintenance
ongoing support is agreed separately, only if you need it.
Architectural Insights
Human-in-the-loop beats raw accuracy
Side-by-side verification reduces errors more than chasing a smarter model.
Treat knowledge like code
Staging, promotion, one-click rollback, and audit trails for every document.
Measure before you believe
Run custom multi-run evaluation suites with at least three runs per question to counter LLM output variance, and guard the evaluation set against biased curation.
Interested in shipping similar systems?
I partner with teams to design, evaluate, and ship production-grade AI systems with verified operational impact.
Let's talk on LinkedIn