Javier OlivieriForward-Deployed AI EngineerForward-Deployed AI Engineer · Software + applied AI
Real problems.
AI solutions that work.
I work directly with your team to uncover bottlenecks and build reliable software enhanced by AI only where it creates measurable business value.
- AI where it adds value
- Deterministic code where predictability matters
- Controlled operating costs
Based in Rosario, Argentina. Working worldwide. · English / EspañolBased in Rosario, Argentina. Working worldwide.
- TODAY
- Agentic AI Engineer at PeopleScape.AI
- FOUNDER
- DevLabs, an applied AI consultancy
- TRACK RECORD
- 17+ years in software · 6+ leading teams

Javier Olivieri · Forward-Deployed AI EngineerRosario, Argentina
From problem to outcome
- 01Business problemManual work, scattered data, delays.
- 02Discovery with your teamI map the real process and where it hurts.
- 03A system built for youCode for rules and integrations. AI only where interpretation is needed.
- 04Business outcomeMeasured in time, cost and errors.
Sound familiar?
You don't need to know AI to start. You need to know where work gets stuck.
Repetitive manual work
Copying data, building quotes, answering the same questions every day.
Costs that grow with volume
Every new order takes more hours or more people.
Decisions on scattered data
Spreadsheets, email, CRM and PDFs that don't talk to each other.
Processes that don't scale
What works with 10 customers breaks at 100.
Slow sales and support
Unanswered inquiries, leads going cold, slow quotes.
Operations full of exceptions
Edge cases only one person knows how to handle.
What your business gains
Less manual work. More focus on the business.
I don't sell technology. I help your operation do more with the team you already have.
Fewer repetitive tasks
The system prepares what's copied, calculated and answered by hand today.
More productivity
Your team spends its time selling, serving customers and deciding.
Costs under control
AI only where it pays for itself. Everything else, efficient and measured code.
Better decisions
Data that's scattered today, organized and ready when you need it.
What I do
Four ways to bring AI into your operations, on code that holds up.
AI agents for business
Agents that read, query your systems and prepare answers or proposals for review.
e.g. order intake and quoting
See service →Process automation
Workflows that move data and tasks between tools without manual work.
e.g. from form to CRM to report
See service →AI assessment and roadmap
What to do first, what to skip, and what each initiative can return.
e.g. prioritize 3 use cases
See service →Integration and measurement
I connect CRM, ERP, Drive, Sheets, APIs and databases, and measure what changes.
e.g. a dashboard of time saved
See service →Hybrid architecture · cost-aware AI
Intelligence where it matters. Solid engineering everywhere else.
I don't add AI by default. I use it where interpreting complex information or making contextual decisions benefits from it. For calculations, business rules and integrations, I choose predictable, auditable, efficient code.
- 1 · InputA task arrivesEmail, form, drawing, PDF or conversation.
- 2 · Rules & routingWhat does it need?Code decides whether it needs interpretation or a rule is enough.
- 3 · Selective AIInterprets the ambiguousWhy here: it needs to understand text, documents or images a rule cannot read.
- 4 · Deterministic validationBusiness rulesCalculations, limits and checks. Always.↳ Human review when confidence is low or the decision is sensitive
- 5 · OutcomeReady to useA quote, a reply, a record or a report.
- No AI required → straight to validation
Path: input → rules → selective AI → validation → outcome
For what must always come out the same
Pricing, calculations, policies, validations, states, permissions and data sync.
Best for stable, auditable rules.
For what needs understanding
Text, documents, images, intent, semantic classification and ambiguous information.
Useful where rigid rules fall short.
For what shouldn't be delegated
Critical decisions, approvals, exceptions and low-confidence cases.
People decide; the system prepares.
Real example · quoting for a stone fabricator
What AI did, and what code did
Interprets drawings and the customer conversation.
Applies business rules, calculations and process checks.
Generates the quote as PDF and Excel.
AI steps in only where understanding is needed. Anything that must be exact is handled by code.See the case →
Before adding a model, I weigh
The goal is net operating benefit, not more or less AI.
- cost per operation
- time saved
- reliability
- latency
- maintenance
- exceptions
- governance
“The best solution isn't the one using the most AI. It's the one solving the problem best.”
Forward-deployed method
I work inside the problem, not from the outside.
I don't just recommend AI tools. I work inside the problem with your team, build a practical solution and help put it into real use.
- 01
Discover
Sessions with owners and users. I observe the real process and its numbers.
- 02
Prioritize
We pick what pays off most. Including deciding where not to use AI.
- 03
Build
A prototype when it helps, then production code integrated with your systems.
- 04
Operate and improve
Measurement, tracing and tuning with your team until it's adopted.
Industries
Where AI already solves concrete problems.
Human resources
Workforce intelligence, team analytics and reports that build themselves.
IMPLEMENTED CASESales and e-commerce
Lead intake, fast replies and quotes ready to send.
POTENTIAL SCENARIOSOperations and finance
Receipts, reconciliations and dashboards from data that's scattered today.
POTENTIAL SCENARIOSLogistics and distribution
Orders, tracking and exceptions that are handled over chat today.
Work and credibility
Real work, with its real status.
- IMPLEMENTED CASEin production for a client
- WORK EXPERIENCEas part of a team
- PLATFORMown product
- PROTOTYPEproof of concept
D'Angelo: from floor plans to quotes, without manual work
An agent that reads floor plans, applies business rules, checks stock and generates the quote as PDF and Excel.
- Claude API
- Vision
- Next.js
- TypeScript
- Google Sheets
PeopleScape.AI
I work as an Agentic AI Engineer on a people analytics and workforce intelligence platform, building and maintaining multi-agent workflows with retrieval, structured outputs, evaluation and observability.
- Claude API
- LangGraph
- RAG
- Multi-agent
Product and brand owned by PeopleScape.AI. I contribute as part of the team.
OWN CONSULTANCY · FOUNDERDevLabs
My applied AI consultancy: client discovery, architecture and development of agents, automations and integrations.
- Discovery
- Agents
- Automation
- Integrations
Valentina
D'Angelo's chatbot that collects measurements and materials and quotes in real time.
AI FOR · streamed conversation
- Claude API
- Streaming
- Next.js
Bold Finance
A multi-tenant finance platform for a café and an academy.
AI FOR · natural-language logging and receipt reading
- Next.js
- Claude API
- Supabase
HealthHue
Voice dialer for a US healthcare client, with handoff to an operator.
AI FOR · voice conversation
- Retell AI
- n8n
- React
Products, brands and names belong to their respective owners. My role was as an engineer and developer. Illustrations are my own schematics.

About
Hi, I'm Javier. A software engineer who takes AI to production.
I've been building software for more than 17 years across fintech, healthcare and global digital products (Securitize, Nifty Gateway, Sama Fertility), and I've led engineering teams for over six years.
Today I work as an Agentic AI Engineer at PeopleScape.AI and, through DevLabs, help companies from the first meeting to a working system. I talk with founders, managers and operations teams, and I write the code.
How I work, in five principles
The right solution, not the trendy one
If a rule does it better, I use a rule.
People where it matters
Human review on sensitive decisions.
Fits what you already use
No forced switch of tools.
Secure and observable
Tracing, limits and permissions from day one.
Measurable results
We define what to measure before we build.
FAQ
What people usually ask me
Something else? Let's talk it through on a call.
No. Many tasks are better solved with traditional code: it is cheaper, faster and easier to audit. I use AI where ambiguous information needs interpreting and combine it with rules, validations and human review.
Yes. I work remotely with companies in the US, Europe, Latin America and beyond, in English or Spanish. I can travel for on-site meetings when the project needs it.
You walk me through the problem, how you handle it today and what you want to achieve. I will tell you honestly whether AI makes sense and what a reasonable first step would be.
It depends on scope. After the assessment I propose a tailored solution with clear phases, deliverables and costs, including the cost of running the AI.
When the case calls for it, yes: CRM, ERP, Google Drive, Sheets, PDFs, forms, APIs, databases and in-house systems.
With validation after every AI step, usage limits, per-operation cost tracking, tracing and human review when confidence is low.
What problem would you like to solve with AI?
In 30 minutes we'll see whether AI makes sense for your case. And if it doesn't, I'll tell you that too.
You talk to me directly,not to a salesperson.