I build the AI features that make it to production.
I'm Aniket — an independent engineer who designs and ships real systems, not demos. I built an entire live AI SaaS solo: retrieval chat, multi-step LLM pipelines, and the infrastructure under them. And when a product needs 3D in the browser, I build that too.
An AI business-strategy platform, built solo.
Vizologi helps founders and analysts reason about business models with AI. I designed and built the whole product — front end, back end, infrastructure, and every AI feature in it.
Three AI features, one engineer, running in production.
Each of these is a distinct pattern I designed, built and keep running against real usage — with the cost control, queuing and failure handling that separate a feature that ships from one that stalls in a notebook.
Ask in plain language
A question about companies or markets pulls the right records from the database and answers with generated charts and cited sources.
Autopilot pipeline
One short prompt fans out through a sequenced, queued pipeline to produce a full business model canvas, SWOT and PEST analysis.
Instant company breakdown
From just a company name and URL, generate a structured business-model overview in seconds — bounded to stay useful, not generic.
Under the hood — Node.js + MongoDB Atlas · asynchronous job queues for the multi-step pipelines · source-grounded retrieval · and model/token-cost tuning worked out against real production limits (smaller models that finish structured output where larger ones truncate). Currently extending retrieval from keyword to vector/hybrid search.
Real-time product configurators, in the browser.
Interactive 3D for physical products — customise a product and watch it update live. Built with Three.js on the front end and Blender Pythonscripting to generate the geometry. Few AI engineers ship this; I've shipped three.
- 01 Ice-pop configurator Three.js · Blender
- 02 Chocolate configurator Three.js · Blender
- 03 Shoe-sole configurator Parametric geometry
replace this placeholder block ]
What I actually do.
Retrieval-augmented chat, structured LLM outputs, multi-step generation pipelines — built to run reliably in a live product, not a one-off demo.
Wiring OpenAI and Claude into existing backends: queues, cost control, failure handling — the parts that decide whether AI ships or stalls.
The full stack under the AI: APIs, MongoDB data models, deployment and monitoring. Fifteen years of keeping systems up.
Interactive configurators with Three.js and Blender — real-time geometry in the browser, generated from parametric models.
One experienced engineer, owning the whole problem.
I work independently and take focused, well-scoped engagements — the kind of work where a single engineer who has shipped every layer beats a larger team still finding its footing.
Based in Goa, India, I've built and run production software with teams across Europe and the US for over a decade. I care about the boring parts — the queue that doesn't drop jobs, the cost that stays predictable, the deploy that comes back up.