Orpita Das
Data Engineer, ex-Snowflake. I build data pipelines that keep working when the world doesn't — and lately I build them with AI.
What I'm good at, and where to check
Noëtone — see it working
Built in 3 days · hardened over months of daily useGold Independently assessed on the Kanz AI Portfolio — certificate KANZ-ADV-1601. “Resilient AI Pipeline: turning scans and handwriting into audiobooks.”
What that means: awarded from 3,500 submissions to the July 2026 Kanz hackathon, judged on originality (20%), technical depth (30%), impact (30%) and presentation (20%) — 60% of the score on depth and impact.
Built solo. Architecture and priorities mine, implementation AI-accelerated. The OCR prompt is a config parameter: other readers narrate a chart's broken axis labels; this one says “the chart shows a declining trend”, because the prompt told it to.
catadoption_pune — the pipeline behind the board
Founded 2024 · Pune · Instagram-first
Three years of Instagram DMs, turned into a live adoption board.
Rescuers send me kittens by DM — photos, an age, a phone number, buried in a thread among thousands. That pile is now a pipeline. Here is the whole of it:
The yellow box is the design decision. The model never publishes — it prepares a case and a human approves it. Automation that can act on its own is automation you have to supervise forever, which is not a saving. I built it to run once, without me, and it published something it shouldn't have; the gate is what came out of that.
Built with Claude — the pipeline, the judgement stage and this page. I decide what it should do and check what it did.
Things I've built
Receipts to a ledger
mindful-spend
Turn a payment screenshot into a categorised expense, with a dashboard over the result. Your spending data never leaves your machine.
Flask · Gemini Vision · Plotly · CSV
github.com/odas/mindful-spend ↗Interview prep that keeps score
Skill Refresh
A quiz and spaced-repetition app for technical-interview prep. Questions adapt as you improve; progress is tracked. Built in three days, live on Render.
Python · Streamlit · Gemini
Open the app ↗Wall art that's actually useful
Kobo Sky Dashboard
A live night-sky map — stars, planets, moon phase, weather and AQI for Pune — rendered as one page for an old Kobo e-reader mounted on a wall. Runs on the home network, no API keys.
Flask · PyEphem · SVG
github.com/odas/sky-kobo-dashboard ↗Building next 🧬 NCBI Taxonomy Explorer — ask “which genus has the most species named after people?” in plain English, over the public taxonomy database.
Background
Independent developer — Noëtone
Pune
A document-to-audio pipeline on a free API tier, rated GOLD out of 3,500 in the 2026 Kanz Build-with-AI hackathon.
Founder — catadoption_pune
Pune
A data-driven cat-adoption initiative on Instagram: ~120 street and Indie cats rehomed; the DM pipeline and public board above.
Cloud Engineer
Algorithmic Biologics, Bengaluru
Cut storage cost ~50% (EBS→S3); a reusable EC2 toolkit for the team's genomics jobs; deployed a gene-expression app to AWS. AWS, GCP, Julia, Python, Docker.
Data Engineer
Snowflake, Pune
Real-time SOX-compliance monitoring built with Internal Audit, saving 100+ manual audit hours a quarter; 20+ ELT pipelines (GitHub GraphQL, ServiceNow, Salesforce, Okta and more); seven external talks including IIA San Jose. Among the Pune office's first hires.
Machine Learning Intern
University of Waterloo
Ensemble QSAR model over protein features, Bayesian tuning, Dask.
Research, teaching and lab roles
UBC · UofT · Natural Resources Canada · Stem Shock (Vancouver biotech startup)
Conifer cold-tolerance proteins; a programmable RNAi herbicide at an early-stage startup; a computational-biology project on microbial SNPs.
M.Sc., Plant Molecular Biology — UBC (degree conferred in Botany) · B.Tech, Biotechnology — NIT Warangal
SnowPro Core, Snowflake, 2023–2025
Hiring for a data or AI-tooling role?
The résumé is the short version and the repos are the long one. I'm in Pune, open to Bengaluru, and I answer email.