Ship AI features in weeks,without hiring an AI team.
I build production RAG assistants for SaaS products and agencies — documentation search, support deflection, internal knowledge. Founder-led: I write the code myself.
13,000+ conversations processed in production · Live on Google Play · Italian + English
AlfaCura — Italian healthcare AI · DocsChat · DocSphere
Why these don't hallucinate
Deterministic rules run before the model, not after.
- 01
Regex emergency detection
Red-flag cases are caught by deterministic rules and never reach an LLM.
- 02
Rule-based routing
Known intents follow deterministic paths — no inference, no guessing.
- 03
Italian entity extraction
Clinical entities are extracted as structured data, not inferred by the model.
- 04
Stateful dialog manager
Conversation context is held outside the prompt, so it can’t drift.
- 05
Schema-constrained output
The model cannot return malformed data — the schema rejects it.
This is the pipeline behind 13,000+ production conversations in Italian and English — the architecture behind AlfaCura.
Fixed-price outcomes
Scoped deliverables, priced up front — not a day rate you have to compare hourly.
RAG Pilot
A working assistant on your own docs plus an accuracy report on 100 real tickets. Credited to the full build if you proceed.
Start with a pilotSupport Deflection System
The complete production RAG assistant across your documentation — deployed, embedded, and answering.
Explore RAG assistantsAI Care Retainer
Ongoing monitoring, prompt and index tuning, model updates, and a monthly performance report.
Explore RAG assistantsNot sure it's worth it? Calculate your deflection savings →
Running an agency? See white-label engagements →
Founder-led · Rome, Italy
Hi, I'm Ojas Gangwal.
Khaas — also known as Khaas Tech and Khaas Agency — is my one-person development studio. I build production RAG assistants and full-stack AI features for SaaS founders and agencies.
Every project I take on, I write the code myself. No offshore team. No junior dev sandbagging your timeline. No account manager between you and the engineer building your product.
Featured Work
Live in production — AI healthcare, RAG SaaS, clinical practice management, and interactive 3D commerce.
What clients say
Ojas built almost the entire AlfaCura platform on his own — the AI triage, the physio matching, the whole practice suite. What stayed with me is that he worried about the parts I hadn’t thought of. The safety layer that catches an emergency before the AI ever sees it was his idea, not something I asked for. He treated our patients’ data like it was his own. I’ve worked with agencies before; this was the first time the person building it clearly cared whether it actually worked.
Working with a Rome-based studio from the US
Three things US founders want to know before the first call.
Timezone overlap, stated
Rome-based — 16:00–20:00 CET overlaps 10:00–14:00 ET. Your morning is my afternoon.
Clean invoicing
Invoiced under an EU VAT number (Partita IVA). Pay by bank transfer or card.
English-native, async-first
Clear written English, documented decisions, no meetings you didn’t need.
Questions founders ask
Straight answers, no sales pitch.
What does Khaas build?
Production RAG assistants — documentation search, support deflection, and internal knowledge tools — plus full-stack AI features for SaaS products and agencies. Founder-led, with 13,000+ conversations processed in production.
Who actually writes the code?
Ojas Gangwal, the founder. No offshore team, no junior devs, no account manager between you and the engineer building your product.
Why don't your AI systems hallucinate?
Deterministic rules run before the model, not after: regex emergency detection, rule-based routing, entity extraction, a stateful dialog manager, and schema-constrained output. That pipeline is behind 13,000+ production conversations in Italian and English.
How much does a RAG assistant cost?
Most teams start with a €6,500 RAG Pilot — a working assistant on your own docs plus an accuracy report in 3–4 weeks, credited to the full build. A complete Support Deflection System is from €18,000, and ongoing AI Care is from €1,500/month.
Are you based in the EU? Do you work with US clients?
Rome, Italy. 16:00–20:00 CET overlaps 10:00–14:00 ET, invoicing is under an EU VAT number, and all communication is in native English.
Ready to Start Your Project?
Let's scope it together — no obligation, no sales pitch.
Book a 30-min call