Fullstack AIengineer &builder.
I build things that scale — from GenAI workflows to full-stack products. AI first, always.
Things I'vebuilt & shipped
01Melina Studio
Scoped LLM memory per board — context isolation across concurrent canvas sessions.
A Figma-like infinite creative canvas with an AI copilot capable of understanding board context, selections, and performing live canvas edits through natural language. Built real-time AI-assisted editing, scoped memory per board, chat-based edit history, and intelligent canvas interactions.
ATREAN (Data + AI SaaS)
ClickHouse for OLAP + Pinecone for semantic — two databases, two jobs, neither compromised.
Built a data-to-AI analytics platform where organizations can upload structured and unstructured datasets, interact using natural language, generate reports, and run decision intelligence workflows powered by LLMs and analytical frameworks.
Spark (Dating App)
SQS for persistence, Redis pub/sub for speed — decoupled delivery without losing messages.
Built entire messaging/chat system when i was in Hushl (Atrean now) from scratch with AWS SQS, Redis, Socket.IO, and MongoDB. Designed real-time message flow, offline support & push notifications.
DeliveryPlus (Logistics SaaS)
OCR pipeline processing WhatsApp-forwarded receipts at 2000+ agent scale.
Backend for a COD tracking system used by 2000+ agents in Hushl (Atrean now). Integrated OCR (Tecstract) to auto-extract data from receipts shared via WhatsApp forms.
Who I actually am
3+
years building
10+
projects shipped
∞
hours dedicated
3
PRODUCTS IN PRODUCTION
Fullstack engineer obsessed with building things that actually scale. I work at the intersection of distributed systems and AI — mostly thinking about how to make LLMs useful in production, not just in demos.
How I buildsystems.
Drawn by hand,built in code.
Films that liverent-freein my heart.
Thinking outloud.
How to Run 'Any' System-Level Library on AWS Lambda
Packaging native binaries and shared libraries into Lambda layers to bypass runtime restrictions.
Airbyte Raw Data Transformation Using DBT (Dynamically)
Automating dbt model generation from Airbyte raw tables — no manual schema mapping.
Postgres → ClickHouse: ELT Pipeline with Airbyte and DBT
End-to-end pipeline from relational OLTP to columnar OLAP — ingestion, transformation, query.
Async/Await Isn't Enough: Solving Synchronous Bottlenecks in FastAPI
Why async alone doesn't fix CPU-bound work in FastAPI — and how threadpools and background tasks do.
Running SQL Queries Simultaneously on ClickHouse with clickhouse-driver
Parallelising analytical queries against ClickHouse using Python's threading and clickhouse-driver.
What I'm doingright now.
Lo-fi Radio
While coding · always
Severance S2
Ep 6 · no spoilers
DDIA — Kleppmann
Ch.9 · third read
UI systems
Can't build what I can't design
Bengaluru, India.
Open to remote