MACTAI uses Vision LLMs to extract structured data from handwritten MACT petitions, police accident-scene sketches, and the medical certificates and supporting documents that accompany every motor accident claim file. Built for motor insurers and their legal teams.
A single motor accident claim file filed at a tribunal contains a handwritten petition, a police FIR and scene sketch, medical certificates in mixed scripts, and the vehicle and driving-licence copies. Somewhere downstream, a clerk retypes all of it into the insurer's core claims system.
It is slow, it is expensive, and it breaks whenever a petition deviates from the template the OCR vendor was trained on — which is every time, because every advocate's clerk hand-writes them differently.
We built MACTAI for the parts of this workflow that everyone has quietly given up on automating.
Scanned, faxed, photographed, or printed from a decade-old claims management system — we take the petition bundle you've got and return JSON your systems can use. No templates. No retraining when the next advocate's clerk writes in a different hand.
FORM-G applications, typed and handwritten. Approximately 150 fields across quantum claimed, heads of loss, dependents, vehicle, and policy.
Hand-drawn scene sketches by the investigating officer. Vehicle labels, arrows, point of impact, road geometry.
Station diary entries, IPC sections, witness statements. Regional-script registers parsed without a template.
Hospital records, discharge summaries, disability percentages, treatment bills — the evidence quantum is built on.
Script-aware field binding. "DOB" next to "15/08/1982" is recognised as date of birth, even when the form prints it as "जन्म तिथि".
That is not what a MACT claim file looks like.
Our pipeline reads the petition the way a claims officer would. It follows arrows in the accident sketch. It handles cross-outs and margin notes. It knows that "DOB" next to "15/08/1982" means date of birth, even when the form prints it as "जन्म तिथि".
↳ This is what sits behind our patent filing.
Confidence scores are attached to every field so your QC team knows where to look first.
We ship the platform into your AWS, Azure, or private data centre. Documents never leave your perimeter. This is what most of our enterprise customers pick.
Faster to start. We handle the infrastructure. Documents are processed and discarded — nothing is retained.
Our first enterprise customer is a major Indian general insurer, live on a signed Purchase Order. The engagement started as a focused deployment and is rolling into a three-year Master Service Agreement.
If your team is spending clerk-hours retyping MACT petitions, we should talk.
MACTAI is led by its two co-founders. We came at this problem from the engineering side, which is why the product looks the way it does: the hard part is not the UI — it is reading a MACT petition handwritten in mixed scripts and photocopied fourteen times.
Madhava brings decades of insurance industry expertise to MACTAI. He previously served as Head of Insurance (India & Middle East) at Wipro and held senior leadership roles including VP & Sector Head at IBM Daksh, GM–Special Projects at Wipro Technologies, and AVP at Polaris. He began his career in insurance operations with Sedgwick Forbes Middle East and The New India Assurance Co. Ltd.
Rajesh is an experienced AI/ML entrepreneur and engineering leader. He co-founded Indix, which was acquired by Avalara in 2019, where he led the data platform for large-scale product intelligence. Post-acquisition, he served as Senior Director of Engineering at Avalara, focusing on AI-driven tax automation. Previously, he was a tech lead on GoCD at ThoughtWorks. Rajesh earned a Computer Science gold medal from Savitribai Phule Pune University.
If you have a sample file you would like us to run through the pipeline, send it over. One file or a batch — we'll treat it the same way.
