
Hey YC, we’re Rahul, Rob, and Rajat from KorrAI (W22).
TL;DR: We're building AI agents for engineers who sign off on mine safety reports. It connects site records, links evidence to failure modes, and supports critical decisions. Live on 5 sites with $1.7M booked in 2026, we’re raising a seed. Contact me at rahul@korrai.com.
https://youtu.be/dcJfnTOH7Kc?si=Ibel5QTfFRf2NgfM
Every tailings dam has an Engineer of Record. Their name and licence go on the document that says the dam and slope will hold, or that a closed site is safe. If that document is wrong, the liability is theirs. For example, in 2019, a tailings dam at Brumadinho, Brazil, failed and 270 people lost their lives. Engineers from the firm that certified it as stable were among 16 people charged with homicide.
We started in 2021 as a spin-off from a Canadian Space Agency research project and spent four years building AI algorithms for geological and satellite data processing. When we worked with the engineers who prepare review reports, we saw what signing actually takes. They pull together decades of reports, sensor readings, inspections, and surveys provided by paid consultants, contractors, and site teams. All information is in different formats. They synthesize conclusions by hand every cycle. But with the overwhelming amount of data mine sites collect today, making sense of it across 20 different systems is daunting.
There’s a risk of missed, misinterpreted, or mistaken risk signals.
AI is already in mining. It drives haul trucks, tunes processing plants, and watches camera feeds.
But there’s no solution for safety reporting because general AI gives plausible and, at times, hallucinated answers. It lacks an interconnected understanding of a mine site and the nuance from diverse sources. Our proprietary data, Ontologies, and Agents are purpose-built for that.
TRAIL reads the entire site record, enables engineers to link evidence to each failure mode, and supports decisions that matter for the facility. There’s a simple human-AI collaboration to arrive at the final report the engineer can sign.
The site record is connected: reports, monitoring data, inspection notes, satellite and drone surveys, including the scanned paper nobody has opened in years.
Failure modes: evidence linked to the ways each facility can actually fail, so the engineer sees what supports each conclusion and what doesn't.
Cited answers: when the engineer asks questions, TRAIL answers with citations for each piece of evidence. Where no evidence supports a claim, TRAIL says so instead of guessing.
The engineer's sign-off: the engineer reviews the evidence file and signs. TRAIL does the insight hunting; the engineer keeps the judgment.
Founders & The Team
Rahul - I'm the CEO. My first startup was acquired before I graduated. Before KorrAI, I built enterprise IoT systems for the tier-1 Indian heavy infra companies. I took KorrAI from a 2-person research project with the Canadian Space Agency to a team of 20, while also leading sales.
Rob (CRO) - is a geologist who worked in the field for Atlantic Gold and then explored others before moving into remote sensing. He led the ground monitoring systems our mining customers run today.
Rajat (CTO) - has 14 years of experience in software and worked on AR/VR and other cutting-edge tech. He works on geospatial AI and leads our architecture and data infrastructure for TRAIL.
We’re based in Toronto.
TRAIL is live at operating and closing mine sites, 5 in total, in Canada and the US. We have $1.7 million booked this year, mostly recurring, all of it won from consulting budgets rather than software budgets. Our active customers include major mining operators, a global engineering firm, and a tech giant.
We are raising a seed round to hire forward-deployed and evaluation engineers, because we have more in the pipeline than we can provide. If you invest in industrial or vertical AI, or know someone who does, write to me at rahul@korrai.com
If you know someone in a mining company or work at one, we'd love to do a live tour of TRAIL on a real project site.