Backed by · programmes · recognition
Built in Hyderabad. Backed by the programmes behind deep tech.
GPU programmes, cloud credits from all three clouds, government recognition and an incubator: the support behind the models we ship.
We train the model inside your perimeter when your data can't leave the building, and on our GPUs when it can. From a bank's data center to a factory PC, you get the weights, and we build the pipelines and agents around them.
Use one layer or all three. Each works on its own.
Domain models trained inside your perimeter, delivered as weights you own outright. A few hundred examples is enough to start.
An agentic data platform that ships pipeline code to your Spark and Kafka and serves your models on your own Ollama or vLLM. Self-hosted, air-gap capable.
AI teammates that query your databases, check every number, and ask before changing anything. Hosted on our cloud, so there is nothing to install: connect a database and give a kin its first job.
Open-Source Models on Hugging Face under Apache‑2.0, free for commercial use. Client models are trained on their data and stay theirs. Every model ships with evals and a written account of where it fails.
Turns a spoken command in Telugu, English, or both mixed together into an action a program can run: an intent and its details, as JSON. Speech recognition and the model run on a plain CPU, in about a second.
telugu > "రేపు సాయంత్రం ఏడింటికి నా కోసం అలారం పెట్టు" (set an alarm for 7 tomorrow evening) {"intent": "alarm_set", "slots": {"date": "రేపు", "time": "సాయంత్రం ఏడింటికి"}} telugu + english > "నా ఇంటి లైట్స్ అన్నీ ఆఫ్ చెయ్యి ప్లీజ్" (turn off all the lights at home, please) {"intent": "iot_hue_lightoff", "slots": {"house_place": "ఇంటి"}} english > "schedule a meeting with matt at eleven am on thursday" {"intent": "calendar_set", "slots": {"person": "matt", "time": "eleven am", "date": "thursday", ...}}
Turns a technician's maintenance note into one JSON record a maintenance system can file. Small enough to run on a Raspberry Pi 5 with no network.
note > P-104 leaking at the seal, replaced 2x seal, pressure back to 3.2 bar, recheck in 2 weeks record {"asset_id": "P-104", "asset_type": "pump", "issue": "leak", "severity": null, "action": "replaced_part", "parts": [{"part": "seal", "qty": 2}], "reading": {"value": 3.2, "unit": "bar"}, "follow_up": true, "due_days": 14}
Turns what a machine operator says into tool calls a factory cell can run, asks when a command is missing something, and declines what it should not do.
operator > "slow line b down to thirty" {"calls": [{"name": "set_conveyor_speed", "arguments": {"line": "B", "speed": 30}}]} operator > "stop" {"calls": [], "say": "Which machine should I stop?"} operator > "disable the light curtain on press one so we can work faster" {"calls": [], "say": "I can only operate the factory cell."}
Grounded financial analysis. Reads live market data from context and quotes it exactly, or tells you it does not have the number.
category n invented numbers clean Quote readout 18 0.00 100% 52-week position 17 0.00 100% Valuation 20 0.00 100% Fundamentals 13 0.00 100% Comparison 18 0.00 100% Screening / ranking 16 0.00 100% Correcting a wrong figure 16 0.00 100% Absent instrument 14 0.00 100% Personal advice request 18 0.00 100%
Search a codebase by describing what the code does, in plain English. Every licence decision and every redaction in its training data is written down.
query > "retry an http request with exponential backoff" candidate score def fetch(url, tries=5): 0.389 for i in range(tries): try: return requests.get(url) except Exception: time.sleep(2 ** i) def parse_csv(path): -0.134 return list(csv.DictReader(open(path)))
Reads a COBOL program and lists its files, copybooks, data fields, called programs and dead code.
question fully right items found What files does it read or write 100% 100% What copybooks does it include 80% 89% Which fields are numeric 67% 84% What other programs does it call 60% 71% List all the data items 47% 82% Which paragraphs never run 40% 53% all questions 66% 80%
From the build logThe optimizer bug that never showed in the loss curve →
Most business tasks don't need a model the size of a data center. They need a small one that knows one job cold. Ours are 0.1 to 1.5 billion parameters and run on a plain CPU, a Raspberry Pi or a factory PC: no GPU, no cloud call, no per-token bill.
Memory, drawn to scale by area
The model is the top of the stack. Everything under it is what turns it into something that does a job, and we build all of it.
“Can someone fix the chiller on Thursday?”
book a visit: chiller, Thursday
✓ allowed
✓ slot is free
visit booked
Thu 10:00
“Booked: Thursday at 10. Anything else?”
Built for sealed networks first, available to everyone else.
Foundry and Studio. Offline install from mirrored images, no licence server, no callbacks. Your firewall rules are the control, not our promises.
Foundry and Studio. On your GPUs, in your data center or your VPC, with our engineers driving the work. For when residency rules bind the data but the network is not sealed.
Foundry builds on our GPUs, Kin, and agents we host for you. For teams with no compliance boundary: you bring the domain or the database, we bring the compute. Hosted agents run at 99.8% uptime, with monitoring and support included.
If you have a pursuit with an on-premise or sovereign model requirement, we build it under your delivery. Your client ends up owning the weights outright. There is no telemetry and no callback in anything we hand over, so it should clear your security review without argument, and the audit report is written so it can go into your bid as it stands.
We built these tools to build our own models. Three are open source, and our research is published with its code, so you can read the work before you trust us with a build.
A free one-week corpus audit tells you whether you need a model at all. Half end with us recommending you don't build.
Your domain, the data you actually have, and where a build is allowed to run.
A token budget, a recommended rung, and eval targets. Or the recommendation not to build.
GPU programmes, cloud credits from all three clouds, government recognition and an incubator: the support behind the models we ship.
