AI systems shipped
30+
RAG, semantic search, copilots, and agentic workflows moving from prototype to production.
// about
I help commerce & healthcare teams ship agentic AI that reaches production — not slideware.
I'm Pras Vakati. I run vakati-tech.ai as a hands-on lab for agentic AI, RAG, semantic search, and commerce architecture — turning AI ideas into reference architectures, working demos, and guardrails your team can actually ship.
Picking the right model is part of shipping, so the site also tracks every major LLM live — price, context, benchmark scores and what each point of capability really costs — alongside guides on RAG, agent evaluation, MCP and healthcare drug commerce.
Agentic AI architectureHealthcare drug commerceRAG & evaluationLLM selection & costSAP Commerce / HybrisMCP & tool gateways
Not ready to talk? Get the lab notes — a short note when something meaningful ships — or follow new model releases.
Pras Vakati
Founder & Principal · Agentic AI and commerce architecture
Greater Philadelphia
Commerce platform scale
$60B+
Architected and led delivery on a headless SAP Commerce platform powering $60B+ in annual healthcare e-commerce revenue.
Hands-on years
20+
Architecture, engineering leadership, and delivery across healthcare, retail, and global commerce.
From enterprise transformation to hands-on AI architecture.
The focus is pragmatic: define the workflow, map the systems, design the guardrails, and build enough working software to prove value. Whether the problem is agentic checkout, pharmacy refill orchestration, semantic product discovery, or SAP Commerce modernization, the goal is the same: help teams ship something credible and maintainable.
Every engagement leaves artefacts your team can adopt and evolve — reference flows, codebases, evaluation harnesses and delivery plans — rather than a deck.
- Agentic AI strategy, architecture flows, and tool orchestration
- Healthcare drug e-commerce use cases, guardrails, and workflow design
- Vector search, semantic pipelines, RAG, and evaluation harnesses
- Model selection on measured cost and quality, not list price
- Composable commerce migration strategies
- MCP server integration patterns and agent-ready tool gateways
- Leadership enablement: AI literacy, governance, and change management
What this site tracks
Live model data for the decision every AI project starts with: which model, at what cost.
- Models tracked
- 445
- Providers
- 66
- With benchmark scores
- 28
- Coding tools compared
- 7
LLM leaderboard
Every model ranked on benchmark score, price and measured cost per point of capability.
OpenAI coding tools compared
Claude Code, Codex, Cursor, Copilot and more — every plan, the models it includes, and how they score.
OpenLLM cost calculator
Describe a workload and see the monthly bill across every priced model, cache discounts included.
OpenLiveBench 2.0 explained
The new categories, the weighted Overall score, and which models rose and fell.
OpenHow the data works
Published sources only
Prices, context and capabilities from OpenRouter; scores and measured spend from LiveBench; coding-tool plans from each vendor's pricing page.
Refreshed every 15 minutes
New models get a page the day they're callable, and price changes show up without a redeploy.
No invented numbers
No blended scores of our own and no guessed matches. A model without a published run shows blank — “not evaluated”, never “bad”.
Commerce & healthcare AI labs
Architecture diagrams, scenario playbooks, and working demos that connect AI capability to commerce and healthcare workflows.
Agentic AI in E-Commerce
Commerce and healthcare drug e-commerce scenarios with examples, guardrails, and operating models.
OpenAgentic AI Architecture
Architecture flow diagrams for shopping agents, pharmacy workflows, and multi-agent orchestration.
OpenDrug Catalog AI Search
Synthetic drug catalog semantic search with filters, ranking, and explainable product discovery.
OpenCareer
Two decades of engineering rigour, product judgment and leadership — the mix needed to move AI from slides into systems.
2024 — now
Founder & Principal, vakati-tech.ai
Practical AI playbooks across agentic commerce, RAG, architecture patterns, and healthcare drug e-commerce.
2018 — now
Lead Application Architect · Cencora
Architect and engineering lead (previously Dev Lead) on a headless SAP Commerce + React/Node platform powering $60B+ in annual healthcare e-commerce — six storefronts and three mobile apps. Built a health-systems storefront from scratch with 40+ engineers and prototyped a virtual-assistant chatbot on AWS.
2017 — 2018
Senior Application Consultant · UnitedHealth Group
COMPAS product development for one of the largest healthcare organizations in the US.
2011 — 2017
Senior Software Engineer · eBay
Lead developer on the commerce services team — e-commerce service solutions, platform architecture, performance, and mentoring offshore engineering teams.
2008 — 2011
Senior Java Developer Consultant · Vanguard
Agile delivery on Vanguard Investment Group’s rapid-response team.
2005 — 2008
Technical Lead / Architect · Citi
Designed and built web-based automation tools for Citi Cards and led a team of seven.
Shipping an AI feature or choosing a model?
Architecture reviews, model selection on measured cost, RAG and agent evaluation, and SAP Commerce modernisation.