Pick up the phone.Read the file.Leave an audit trail.
Solaich Labs builds production AI for Indian businesses and enterprise teams — voice, documents, and the governance layer that keeps them safe.
Five things running in production and pilot.
ALCP — Agent Learning Control Plane
The governance layer for agents that take real actions. Risk and ops teams write rules in plain language; the SDK enforces them in-process, so enforcement holds even if the control plane is offline. Every decision lands in a hash-chained, tamper-evident audit log.
Catloop
Digital catalog for importers and wholesalers. Hinglish voice search, Tally sync, and public catalog pages structured so ChatGPT, Perplexity and Gemini can cite your products by name.
Voice Agents
An assistant that answers the calls and messages your team doesn't get to — quality complaints, order queries, appointment booking, payment follow-ups. Hindi and English. Embeds on your site, catalog page or WhatsApp.
CaseDesk
Document search for advocates. Upload judgments, contracts and case files; ask in plain language and get the answer with the paragraph it came from.
BoothBrain
Expo lead capture without the paper forms. QR at the booth → visitor speaks → transcribed, structured and in your reps' hands before they leave the stall.
“Answer questions from my policies”
You have 400 policy PDFs, a team that keeps asking the same 30 questions, and no appetite for an assistant that confidently invents a clause. Here's exactly how we build it.
{{ s.body }}
What usually breaks{{ s.breaks }}
“Answer the calls we're missing”
{{ s.body }}
Seventeen years of enterprise systems before any of this was called GenAI.
EDI X12 837, 270/271, claims and encounter pipelines, payer platform data models. The unglamorous side of healthcare IT where a schema change breaks a claims run.
Azure AI Foundry, Semantic Kernel, Microsoft Agent Framework, MCP servers over SQL Server, GitHub, Jira and SharePoint — long-running agent workflows that do discovery, analysis, remediation and raise the pull request.
DeepEval for regression, Presidio for PII, hash-chained audit for anything that touches money or coverage.
Python/FastAPI, Node, PostgreSQL + pgvector, MS SQL Server, multi-LLM routing across OpenAI, Claude and Gemini with cost and fallback logic.
How an engagement runs.
We map the use case, the data, and whether AI is even the right answer. You leave with a written scope whether or not we work together.
We score your data, process, volume, risk tolerance and ownership, then put a number on what each candidate use case returns in hours or rupees. You get a ranked shortlist, the honest not-yet list, and a cost model. Fixed fee, credited against the pilot.
Your real data, your real questions, a working thing you can put in front of your team. Fixed price.
Evals, access control, audit logging, deployment, handover docs.
Monitoring, retrieval tuning, model and cost review. Cancellable.
Solaich Labs is led by Gurpreet Singh Solaich — 17 years in enterprise systems, the last three building agent platforms. Engagements are delivered by a small vetted bench of engineers assembled per project. You talk to the person who designs and writes the system, not an account manager.
The questions founders actually ask.
{{ q.body }}
Tell me what you're trying to build.
A 45-minute build review, free. We map the use case, the data, and whether AI is even the right answer. You leave with a written scope whether or not we work together.
Two minutes, and only the name and a way to reach you are required. I read every one myself and reply within a working day — or tell you honestly that it isn't a fit.