Built on Data.
Run by Intelligence.
Most organizations are sitting on data that should be driving decisions. Yours doesn't have to be one of them.
SWIGGY · REAL-TIME ON FABRIC
1B+ events a day
Streaming decisions at India scale on Microsoft Fabric.
GRASIM · FIBRE QUALITY CONTROL
4 hrs → 1 min
Defect detection at 98% accuracy, covered in Forbes.
AP COMMERCIAL TAXES · AI TAX OFFICER
21% net GST growth
Department-reported, with AI-based scrutiny one driver among several.
SEAIR EXIM · MICROSOFT IGNITE 2024
90% faster data processing
Shown in Satya Nadella's Microsoft Ignite 2024 keynote.
Quality caught in-process,
not after the batch.
AI-vision fibre inspection at the edge of the line: 10,000 images a day at 98 percent defect-detection accuracy, and 40 percent of material cost saved because defects surface while the batch can still be corrected. As covered in Forbes.
Read the full storyWhat we do.
One accountable team, from strategy to a tuned estate.
Judgment before the build. We map what your data can actually do for the business, price the path, and hand you a plan you keep, whichever way you decide.
We move your data out of aging systems onto a foundation built for what comes next, without losing a single number along the way.
The work your teams feel every day: live operational intelligence, reporting people trust, and AI that acts inside the systems you already run.
A running estate quietly gathers waste. We find it, tune it, and turn the savings into the budget for whatever you build next.
Trusted by Leading Organizations.
Recent insights.
View all insights
Claude Workflows Got the Architecture Right. The Bill Is the Open Question.
Anthropic moved agent orchestration out of the model's context window and into JavaScript you can read, the same pattern we've hand-rolled in LangGraph for two years. What's genuinely new, what we found running it on our own code, and the one number that should make you read the docs before you type "ultracode."
Building a Construction SLM
Field notes on building a small language model for construction text: the classification problem that broke a frontier API, the seven-billion-parameter base that fixed it, and the hybrid architecture that ships in production today.
On-Prem AI Is Coming Back. The Hyperscalers Will Pretend It Isn't.
Sovereignty has gone from a slide to a contract clause. Open-weight models have closed the gap. CFOs have finally seen the bill. Three forces pulling enterprise AI back inside the firewall, and what the hyperscalers will do next.
Mandelbulb Showcases AI Tax Officer at the India AI Impact Summit 2026
Mandelbulb Technologies showcased its AI Tax Officer at the India AI Impact Summit 2026, presenting alongside Microsoft and the Government of Andhra Pradesh at Bharat Mandapam.
Featured on Microsoft: How Mandelbulb Helped Swiggy Deploy Real-Time Intelligence on Microsoft Fabric
Mandelbulb Technologies' Co-Founder Deepanshu Thakur is quoted in an official Microsoft feature on Swiggy's deployment of Real-Time Intelligence on Microsoft Fabric, which cut operational latency from 10 minutes to seconds.
Mandelbulb Opens Its Singapore Office
Our first overseas office and the first dedicated home for our Southeast Asia engineering and customer practice. Why Singapore, who is on the starting team, and what is already in pipeline.
RAG vs Graph vs Hybrid for Industrial Knowledge
Field notes from three Plant Brain deployments on when vector RAG wins, when graph retrieval wins, when the time-series layer wins, and when you have to combine all three. The retrieval pattern, not the model, is the question that takes longest to answer.
Transforming Healthcare with Microsoft Fabric: AI-Powered Discharge Summaries
Introduction Healthcare IT environments grapple with massive, siloed data sets, from electronic health records (EHRs) to imaging and...
Start with a working session.
Thirty minutes with a senior engineer. Your estate, your use cases, one shared shortlist. No deck, no junior bench.
Prefer email? hello@mandelbulbtech.com