Odoo ERP · AI automation · Ahmedabad
We implement, migrate and extend Odoo — then wire the AI layer on top. Agents, RAG and LLM pipelines running against your live business data, built by the engineer who scopes them.
Odoo modules shipped
ERP migrations delivered
Projects delivered
In production · ERP agent
- Model
- Guardrail
- Deterministic
- Someone asks in plain English"What did we ship to Surat last week?"
- Model turns it into a queryClaude MCP
- Their role is checked firstBefore anything executes
- ORM runs it — never raw SQLScoped to what they could already see
- Answer in secondsEvery call logged
The model never touches the database.
30+
Systems in production
Still running. Not pilots, not demos.
See the work120+
Odoo modules shipped
15+ ERP migrations, zero records lost.
Odoo work2
Peer-reviewed papers
Published in Elsevier. Check it yourself.
Read the paperFDE
Forward-deployed engineering
One engineer, embedded, start to finish. No account managers.
How we workWhich of these is you?
“Our Odoo is a mess.”
Half-finished modules, workarounds nobody documented, and a team that has quietly gone back to spreadsheets.
Get it stable“We want AI on our own data.”
Not a chatbot bolted on the side — agents and RAG that read your live ERP and are safe to hand to non-technical staff.
See how it works“Too much of this is still manual.”
Invoices retyped, claims keyed by hand, reports rebuilt every month. Work software should have finished hours ago.
See what we automatedSound familiar?
The expensive stuff is rarely the dramatic stuff.
It is the daily manual work, the data nobody can find, and the systems that only one person understands. Here is what we usually find in the first week.
What we do
One partner. Every layer of the stack.
AI on top, platforms in the middle, cloud and security underneath. We own the whole thing so nothing falls between vendors.
The AI layer
Named properly, because the details are the difference.
Six things that separate a system you can put in front of staff from a demo. Each one is from a build that is running now.
Agentic tool-calling
The agent does not get database credentials. It calls a tool, the ORM decides whether that call is allowed, and the result comes back. Model proposes, runtime disposes.
ERP agentRetrieval-augmented generation
Answers grounded in your documents and tables rather than the model’s memory, so a wrong answer is a retrieval bug you can fix — not a hallucination you can only apologise for.
AI solutionsStructured outputs
Every model response is parsed against a schema before anything downstream trusts it. Wrong shape, rejected — it never reaches your database as a half-parsed string.
Email intelligencePrompt-injection hardening
Untrusted text — an email body, a PDF, a form field — is treated as data, never as instructions. We wrote up what this does and does not defend against.
Read the teardownHuman in the loop
Confidence scoring routes the uncertain cases to a person and lets the rest through. Your team reviews the edge cases instead of every row.
OCR engineModel routing
Five providers behind one registry, so a price change, a rate limit or a better model is a config change — not a rewrite and not a lock-in.
Email intelligenceHow it runs
Diagnose. Blueprint. Forge. Scale.
Four stages, one accountable team. Your first measurable win lands inside two weeks, or we tell you why it will not.
01 · Week 0 · Free
Diagnose
We map your stack, find every bottleneck and quantify what it costs you. 45 minutes, written summary, no pitch.
02 · Week 1–2
Blueprint
Our architects design the solution around your business logic: AI workflows, platform modules, cloud, security. You approve every milestone before we build.
03 · Week 2–8
Forge
We build it in staging, you test it, we ship it. First measurable wins typically land within 14 business days.
04 · Ongoing
Scale
Ongoing optimisation, 24/7 monitoring and dedicated support. The system gets smarter as your business grows.
Proof, not promises
Numbers from real deployments.
Projects delivered
Odoo modules shipped
ERP migrations delivered
Developers mentored
Peer-reviewed publications
Inventory planning accuracy
Stock discrepancies
Less manual data entry
Selected work
Built for teams that measure.
NL→SQL
Plain English to live data
100%
Prompt-injection hardened
Conversational AI Agent for Live ERP Data
Every business question needed a developer, and answers arrived days late.
The hard part: Letting non-technical staff query live ERP data without ever giving the model database access.
A natural-language interface over live ERP data. Non-technical staff ask plain-English questions and get ORM-safe, role-gated answers in seconds.
- Claude MCP over live ERP
- Enterprise ERP client
- Shipped 2025
Read case study
100%
Claims processed automatically
80%+
Manual data entry eliminated
Healthcare Insurance Claims Automation
A Canadian health platform was keying every OHIP claim, eligibility check and reconciliation by hand.
The hard part: Three government interfaces — OHIP, MCEDT and EDI — each with its own failure modes, running unattended overnight.
OHIP real-time eligibility checks, MCEDT payment reconciliation and automated XML/EDI batch claims. What took a team days now runs overnight without oversight.
- OHIP + MCEDT
- Runs overnight, unattended
- Healthcare platform
Read case study
0
Manual steps in the credit lifecycle
100%
Audit trail coverage
Line of Credit Lending Platform
A lending book run out of spreadsheets, where one person understood the formulas.
The hard part: Removing every manual step from a credit lifecycle while keeping a complete audit trail on each state change.
Full lending lifecycle built from scratch: Plaid bank verification, Stripe disbursements and repayments, multi-location credit rules, automated statements and interest.
- 2 payment rails
- Lending enterprise
- Shipped 2024
Read case study
“CodeCrafters reduced our order processing time from 4 hours to 8 minutes. The ERP implementation paid for itself in the first quarter.”
COO
3PL logistics operator · India
96% faster processing
“Their AI pipeline replaced 3 full-time data analysts. The reports are more accurate, available instantly, and cost 70% less to produce.”
CTO
Financial services firm
70% reporting cost cut
“We scaled from 50 to 500 daily transactions without hiring a single additional ops employee. CodeCrafters built the infrastructure that made that possible.”
Founder
Multi-brand retail group
10x scale, same team size
“The OCR engine alone saves my team 3 hours a day on invoice entry. Confidence scoring means we only review the edge cases. Everything else is automatic.”
Finance Director
Manufacturing company
80% manual entry eliminated
“ERP migration with zero downtime and not a single data loss. I was expecting 6 months of pain. It took 8 weeks and was completely smooth.”
CTO
eCommerce group
Zero downtime migration
Role and sector shown; names withheld under NDA. Happy to arrange a reference call.
How we work
Research rigour, applied to production systems.
CodeCrafters exists because most enterprise software is fragile, and the reason is almost never the software. It is the depth of thinking behind it. We were built out of years spent inside a major ERP vendor watching rollouts that technically shipped and never actually landed.
8+ AI systems in production
120+ Custom modules shipped
30+ Projects delivered
2 Peer-reviewed papers behind the work
Insights
Field notes from shipping AI.
09 Jun 20266 min
The AI readiness audit: 14 questions before you automate anything
Most AI pilots die before the first prompt. Fourteen questions across data, process, people and risk that tell you if you are ready.
01 Jul 20266 min
Agents vs. workflows: when an LLM belongs in the ops loop
Deterministic workflow, LLM step, or autonomous agent. A decision framework, the failure modes of each, and where the model sat in a real 4h-to-8min automation.
21 Jul 20267 min
From 4 hours to 8 minutes: anatomy of an order-processing automation
A stage-by-stage teardown of a logistics order pipeline: ingest, extract, validate, reconcile, post, the human gate, and what broke in week two.
FAQ
Questions we get on the first call.
Book a 45-minute call with our architects. We map your current stack, find the three biggest bottlenecks and estimate the ROI of fixing them. You get a written summary within 24 hours. No pitch, no commitment.
Three. A free Diagnostic, a fixed-scope, fixed-price Build, or a Scale retainer where we act as your engineering team. Most clients start with a Build and move to a retainer once the first system is live.
Founder-led. Jaimin Shah scopes and architects every engagement and stays on the account. No offshore handoffs, no juniors on your project. Senior engineers end to end.
We are based in Ahmedabad, India (IST, UTC+5:30) and work with clients across North America, the UK, the Gulf and Australia. We overlap with your working hours for standups and respond within 4 business hours.
How we de-risk it
What you are actually agreeing to.
A diagnostic before a proposal
We map the bottleneck and tell you honestly if AI is the wrong tool. Often it is.
Fixed scope, fixed price
Builds are priced after scoping, not billed by the hour against an open clock.
You keep the code
Full source, full documentation. No proprietary runtime you cannot leave.
The same engineer throughout
Whoever scopes your system builds it and stays on it. No junior handover.
Free download
The AI Readiness Checklist for CTOs.
14 checkpoints we run before any automation engagement. Ten minutes to fill in. It tells you where the money is, and where the risk is.
- Where your data actually lives (and who can read it)
- Which processes are automatable in 30 days
- The three failure modes that kill AI pilots
- A scoring sheet you can hand to your board
Next step
Your competitors are already automating.
A free 45-minute diagnostic. We map your biggest bottleneck, estimate the ROI, and tell you honestly whether AI belongs there. No pitch deck.
Straight to the engineer. No form, no gatekeeper.
Accepting new engagements · Q4 2026