about
I build systems that remember. The reason starts in Varanasi.
I'm a data and machine-learning engineer. In early 2022, after two years of watching the train booking come back waitlisted, I gave up on trains and rode buses for days across four states — Hyderabad, Nagpur, Madhya Pradesh, Prayagraj — and a last train into Varanasi. I sat at the Manikarnika burning ghat past midnight, three nights running, and came home with a question I still haven't put down: what survives the burning?
What survives the burning
That question didn't stay at the river. Some evenings I read a chapter on Aghora and a chapter on a data warehouse, and the seam between them disappears — both are asking the same thing. For years I built systems that moved data reliably, at scale. Over time the question got larger: how does memory actually accumulate, and what makes an intelligence worth trusting? sarpOS is where the river and the pipeline meet.
What I do today
Day to day, I build data and machine-learning systems — the pipelines that move real-world data and the models built on top of them. Most of what's public here is reference builds: complete, runnable implementations on GitHub rather than client work I can show. I'm building sarpOS in the open, and it's early — today that mostly means Vasuki, the part you can talk to. I'd rather show you what's real and name what isn't than oversell it.
sarpOS — the long game
sarpOS is my attempt at a genuinely personal AI operating system — not a wrapper around a chatbot, but a system meant to learn, remember, and act within boundaries I set, on hardware I control. It's early and in development, and it's the throughline of everything else here: every pipeline and model is practice for building a memory I can trust. Vasuki is its voice — the interface to sarpOS you can talk to on this site. It answers from what's public here, and it says so when the record doesn't have an answer.
Tools I work with
- Apache Spark
- PySpark
- dbt
- Airflow
- Kafka
- Delta Lake
- Azure
- AWS
- Databricks
- Snowflake
- PyTorch
- LangChain
- HuggingFace
- RAG
- LLMs
- Vector DBs
- MLflow
- Python
- SQL
- TypeScript
Work with me
I'm open to collaborators on data infrastructure, retrieval, ML, and AI systems — and glad to talk with anyone building something honest. Tell me what you're working on.