Back to FAQ Hub
FAQ Category

Technical Approach

Architecture choices, model routing, RAG vs fine-tuning, eval framework, observability, and security posture for production AI in 2026.

Overview

Technical due diligence is where production AI separates from demos. This page explains how DevStudio chooses orchestration per workflow, when RAG beats fine-tuning, the three layers used to prevent hallucination, and how observability and token audits keep a deployed system accurate and economical. Every choice is documented in an architecture decision record you own.

Key things to know about the technical approach

Orchestration is chosen per workflow

LangGraph is the default for stateful, branching workflows because the state machine is auditable; lighter custom orchestration is used for mostly-linear flows. We avoid frameworks heavier than the workflow they model.

RAG and fine-tuning solve different problems

RAG handles changing facts with cited sources; fine-tuning handles stable voice and format. Most production systems use both, measured against the eval set rather than chosen by preference.

Hallucination control is layered

Retrieval grounding with source citations, deterministic policy guardrails on every tool call, and full observability so any unsafe behavior is reproducible after the fact.

Frequently Asked Questions

Do you build with LangGraph, AutoGen, CrewAI, or custom orchestration?
We pick per workflow. LangGraph is our default for stateful, branching workflows because the state machine is auditable. We use lighter custom orchestration when the workflow is mostly linear with a small number of tools. We avoid over-frameworked stacks where the framework abstraction is heavier than the workflow it is modeling. The choice is documented in the architecture decision record.
When do you choose RAG vs fine-tuning?
RAG handles the facts (changing knowledge); fine-tuning handles the voice or stable in-domain reasoning patterns. Most production systems use both: RAG for grounded answers with citations, light fine-tuning for tone or format compliance. See /blog/rag-vs-vector-search-vs-llm-fine-tuning/ for the decision framework.
How do you prevent hallucination?
Three layers. Retrieval grounding — the agent cites the source document for every factual claim. Deterministic guardrails — every tool call is checked against a policy ruleset before execution. Full observability — every prompt, retrieved context, tool call, and model response is logged so any unsafe behavior is reproducible after the fact.
What models do you use?
Closed-source frontier models (OpenAI, Anthropic, Google) when accuracy is paramount. Open-source models (Llama, Qwen, Mixtral families) when data sovereignty or unit economics demand it. Embedding models picked per workload, measured against the eval set. Every choice is documented and reviewed quarterly so the system rides the model-quality curve.
Can the system run fully on-premise?
Yes. We have shipped fully self-hosted RAG and agent systems on customer infrastructure with open-source generation models, on-premise vector stores, and on-premise observability. The trade-off is unit economics and the speed of riding the frontier model improvement curve, both explicit in the architecture decision record.
What about token cost in production?
Every project includes a quarterly Token Audit re-evaluating routing, caching, and model selection against the eval set. See /blog/ai-agent-token-cost-audit/ for the methodology. Buyers who skip post-launch token audits silently lose 30-50% margin to unrouted frontier model usage.

Related guides

RAG Development

Production retrieval systems with grounded, cited answers.

 RAG vs Fine-Tuning

The decision framework for grounding vs adaptation.

 Token Cost Audit

How quarterly audits protect production margin.

 Delivery & Handover FAQ

What you own and how model upgrades are handled.

Other FAQ Categories

Cost & Pricing

Project rates, Paid Scoping, payment milestones.

 Timeline & Delivery

5-phase delivery, milestones, post-launch.

 Technical Approach

Architecture, RAG vs fine-tuning, security.

 Team & Communication

Time zones, escalation, async-first delivery.

 Engagement Process

Discovery, Paid Scoping, change requests.

 Delivery & Handover

Source code ownership, 6-month QA, model upgrades.