{{ p.caption }}
{{ p.status }}
{{ p.headline }}
{{ p.body }}
{{ pt }}
AI product studio · IIT Madras
We ship ML systems to production. Ours first.
Jatrix Labs builds and runs its own AI products, then builds the same kind of systems for SMBs and startups: data pipelines, models, and the MLOps that keeps them working.
jatrix / production
LIVE
nosides.ingeststream12,400+/day
nosides.entity-extractmodel99.2% prec.
nosides.clash-detectpipeline50+ src/story
farm.platformmodeldeployed
monitor.driftjobwatching
Systems we built and operate ourselves.
2
AI products live in production
12.4k+
articles processed daily by NoSides
99.2%
entity extraction precision
<24h
founder reply on every enquiry
01 — Built in-house
Our own products, running in production.
Before we build for clients, we build for ourselves. Both products run on the same pipeline, model and monitoring stack we deploy for you.
NoSides
nosides.news ↗Real-time news intelligence. Ingests articles and social posts, tags who said what, and surfaces where sources clash on the same story.
WHO'S IN THE NEWS
FULL MAP ↗
INTERACTIVE DEMO / FICTIONAL NEWS
12.4k+
articles / day
50+
sources per story
24/7
live ingestion
Streaming ingestion
NER
Stance detection
LLM summarisation
Farm Platform
In productionAI platform for farmers. Turns field data into clear, timely decisions on crops, inputs and yield.
—
farmers / acres
—
model accuracy
—
languages
Computer vision
Forecasting
Vernacular NLP
Mobile inference
02 — Showcase
From the lab: interactive product prototypes.
Explore a working farm season and a logistics truck in 3D, and browse our phone-first farm app redesign. All showcases use sample data.
03 — Client work / Retail intelligence
Research that became working software.
A real engagement for a Chandigarh pharmacy: understand demand, diagnose inventory inefficiency, and build a useful tool that fits the accounting system the business already uses.
CASE STUDY 01 / PHARMACY OPERATIONS
Completed client work
From Tally exports to smarter stock decisions.
Sales and purchases were sitting inside a shared Tally ledger for two retail operations. We cleaned and separated the pharmacy data, studied demand across seasons and order types, and surfaced the inventory patterns tying up cash.
RESEARCH
12 months of purchase and sales analysis, seasonality, SKU turnover and K-means order segmentation.
DIAGNOSE
Overstock, slow-moving products, reorder thresholds and supplier concentration.
DELIVER
Python + Streamlit inventory application: upload a Tally stock-summary Excel export, clean the data automatically and inspect inventory health.
Python
Streamlit
Tally Excel imports
Scikit-learn
Inventory analytics
INVENTORY / RESEARCH SNAPSHOTFY 2023–24
4,824
SKUs maintained
64%
SKUs overstocked at March cutoff
79%
orders valued below ₹200
47%
SKUs with only one supplier
STOCK EXPOSURE / MARCH 202464%
NO WINTER UNIT SALES26.5%
FROM ACCOUNTING DATA → INVENTORY ACTION
01 / INPUT
Tally stock Excel
02 / PROCESS
Clean + classify
03 / VIEW
Reorder & stock flags
Historical diagnostic findings, not claimed post-launch improvements. Client identity and raw financial records omitted.
04 — What we build for you
From a week-long automation to a model you own.
{{ t.n }}{{ t.time }}
{{ t.title }}
{{ t.body }}
{{ i }}
05 — MLOps
A model is only useful if it keeps working.
Every system ships with the lifecycle around it: versioned data, reproducible training, automated evaluation, monitored deployment and retraining on drift.
{{ s.n }}
{{ s.title }}
{{ s.body }}
Our edge
AI that works in Indian languages.
Speech-to-text, NLP and voice agents tuned for regional languages, accents and code-mixed speech. The same work powers the vernacular layer of our farm platform.
Discuss a vernacular project →06 — How we work
Audit. Prototype. Production.
{{ p.n }}
{{ p.title }}
{{ p.body }}
Output: {{ p.out }}
Founded at IIT Madras. You work directly with the two founders who write and review every line we ship.
R&D · IIT MadrasHQ · Chandigarh, India
Tell us what you want to automate.
Free 15-minute AI audit. A founder replies within 24 hours.
Request received.
We'll reply to your email within 24 hours.