Agents that take over the busywork
Agents that read the email, check the order, update the system and ask a person only when it matters. Every step is logged. Good for support, operations and back-office teams drowning in repetitive work.
Enterprise experience. AI-native skills.
Many AI developers have never shipped code to a system that moves real money. Many enterprise veterans have never built an AI agent. I have done both.
For 19 years I analysed, built, tested and supported business-critical systems for Costco, Bank of America, Humana and Frontier. Today I bring that discipline to AI agents, RAG assistants and automations that work on a busy Monday, not just in a demo.
agent run
refund-request · #48213
0.0s
- Customer email arrives"I was charged twice for order #48213."
- Looks up the orderIn the business system of record: DB2, SQL Server, ERP.
- AI reads the policyClaude matches the case to the refund rules.
- Code does the mathRefund $240.00, exact to the cent. Not the model's guess.
- A person approvesOver $200, so it waits for a human sign-off.
- Done and loggedCustomer answered, every step in the audit trail.
19+years in enterprise IT
7industries
10+enterprise clients
22AI tools you can try
- Costco Wholesale
- Bank of America
- Frontier Communications
- Humana
- Performance Food Group
- MAPFRE Insurance
- West Marine
- Aspen Dental
- Service Experts
Range
One engineer. Every hat.
Projects rarely fail inside a role. They fail between roles: the analyst who never saw the code, the tester who never met the user. I have done every one of these jobs, so nothing falls through the gaps.
- 01AnalystTurns a vague request into requirements and a design developers can build from.
- 02DeveloperShips the code, from RPG and SQL stored procedures to Java, Python and REST services.
- 03TesterBreaks it before your users do: unit, integration and UAT testing.
- 04Team leadKeeps a team moving: assigned work, reviewed code, ran daily status, trained new people.
- 05Production supportAnswers the 2 a.m. call. On-call support, releases and change management.
Built on
Building with
The journey
Inside the systems that run the business. Since 2006.
- 2006MumbaiCGI · Humana insurance conversion
- 2009NoidaHCL · CWDirect warehouse system
- 2011Richmond, VAHCL · Performance Food Group
- 2019Dallas, TXBank of America · Frontier
- 2021Seattle, WACostco Optical digital store
- NowRichmond, VAAI Engineering Lab · freelance
What I build
Hand me the problem. Get back a working system.
Answers from your own data
Assistants that answer from your documents and data, with citations, and say "I don't know" instead of guessing. Plus pulling clean data out of PDFs, emails and forms.
AI that ships as a product, not a demo
The part most AI demos skip: APIs, integrations, databases, scheduled jobs, Docker and CI. You get a product your team can run, not a notebook.
New life for legacy systems
Your core systems hold years of business rules. I expose them as clean APIs that AI agents can call, and use LLMs to document and modernize the legacy code, starting with IBM i and DB2.
Apps
Not slides. Software you can click.
The lab
Don't take my word for it. Try 22 tools.
Every tool runs on sample data or on your own input, so you can judge the work before we talk. Filter by area or search.
No tool matches that search.
Enterprise results
Trusted with systems where a bug costs real money.
2021 — 2026
System Analyst · Costco Wholesale
One set of business rules behind both the in-store green screen and the online store.
Moved RPG business rules into DB2 SQL stored procedures so that order creation, prescription validation and patient look-up serve both the green screen and the online store, returned as JSON to REST services. Tuned slow SQL with Visual Explain and Index Advisor.
2019 — 2021
System Analyst · Frontier Communications
Years of telecom data purged safely, in the right order, automatically.
Automated a data purge across telecom tables in dependency order, with the rules centralized in stored procedures and tuned with Visual Explain.
2019
System Analyst · Bank of America
Escrow calculations with validation, error handling and a full audit trail.
Built the stored procedures for escrow calculation and account maintenance, with parameter validation, exception handling and audit logging.
2009 — 2014
Senior Software Engineer · HCL Technologies
Kept a national food distributor's warehouse and order systems running, and led the offshore team.
For Performance Food Group and CWDirect: supported the warehouse and order management systems, built interfaces to the external order management system, and led the offshore team's code reviews and daily status.
2006 — 2009
Software Engineer · CGI
Built the tool that moved an insurance system to its new owner.
For Humana: designed and built the conversion tool used to migrate an insurance system to its new owner, wrote its user manual and demonstrated it at the client site.
Also: Aspen Dental (Telliant Systems), MAPFRE Insurance (TalTeam), West Marine (Infogain), Service Experts (NTT DATA Services).
Working with me
No black boxes. No surprises.
- 01
We start with one workflow and a working first milestone within days, then build on it.
- 02
Structured outputs with schema validation. Code, not the model, does the math and enforces the rules.
- 03
A person approves the decisions that cost money. Every step the system takes is logged.
- 04
Tests and recorded evaluation cases, so it still works six months later and you can prove it.
- 05
A written update every day. You own the code and the repository.
Now
- Taking AI agent, RAG and automation projects on Upwork (Rising Talent).
- Growing the AI Engineering Lab: 22 tools, each one runnable.
- Turning Job Hunter into a hosted product.
Questions
Before you reach out.
Why IBM i and AI in one person?
Most AI projects fail where the business data is, not in the model. I spent 19 years keeping IBM i, SQL and Java systems correct under real load. I bring the same discipline to AI: data checked before it reaches a model, a person approving any action that changes a record, and every failure logged.
Can you work in our stack?
Most likely. I have worked in retail, telecom, insurance, healthcare, banking, warehousing and food distribution, with teams onshore and offshore. I learn a codebase by reading it, then ship a small milestone so you can judge the work, not the promise.
How do you price a project?
A fixed price per milestone, agreed before any work starts, so you always know the cost. Hourly works too through Upwork.
How fast can you start?
I reply the same day with an approach and a first milestone. The first milestone is something working you can test, delivered within days.
Contract, full-time or freelance?
All three. Freelance projects go through Upwork. For contract and full-time roles, reach me on LinkedIn or by email.
How do you keep AI from making expensive mistakes?
Structured outputs checked by code, exact math done in code rather than by the model, a person approving anything that moves money, and every step logged. You can see this working in the lab tools.
Have a process that eats your team's week?
Send me two lines about it. You get a same-day reply with an approach, a first milestone and a fixed price. Hiring for a role instead? I'm open to contract and full-time positions too.