---
title: 'Weather Scientist (Numerical Weather Prediction) at Pravāh'
canonical: 'https://feeny.ai/job/weather-scientist-numerical-weather-prediction-pravah-delhi-948bf425b632'
type: 'job'
last_seen: '2026-09-09'
---

# Weather Scientist (Numerical Weather Prediction) at Pravāh

- **Company:** Pravāh
- **Location:** Delhi, India
- **Compensation:** $20k–$40k
- **Employment:** full-time
- **Work type:** onsite
- **Posted:** 2026-05-17
- **Last confirmed live:** 2026-09-09
- **Apply:** https://jobs.ashbyhq.com/pravah/3d580ee8-0b2b-411a-bd39-ca2e8921e5bb

## Job description

## ABOUT PRAVĀH

Pravah is building the world's first foundation model of the electric grid. The world's most important physical industry is under unprecedented stress. We are using ML to build the 22nd-century grid. We are backed by Khosla Ventures, Pear VC, and Conviction.

We work directly with state-owned distribution and transmission electricity utilities to transform their decades of fragmented operational data into deployable, decision-grade intelligence. Our customers include some of the largest DISCOMs in the country, and our work spans demand forecasting (including renewable generation forecasting), weather forecasting, network mapping, and load flow analysis.

## THE ROLE

We are hiring a Weather Data Scientist to advance the next generation of weather forecasting systems for India, with strong attention to observational data quality and geospatial consistency. You will work closely with machine learning and software engineers on two core threads:

- Numerical weather prediction: run regional NWP models to generate high-resolution forecasts and training data.
- Data assimilation: contribute hands-on to data assimilation for weather forecasting models.
- ML-ready datasets: procure, process, and create ML-ready global and regional weather datasets at large scale (high volume, multi-source, long time horizons), with explicit focus on data-sparse regions.

## WHAT YOU'LL WORK ON

- Build and benchmark next-generation multiscale, regional, and global forecasting systems against reanalysis and observations, with particular focus on nowcasting and extreme events. The work rests on careful treatment of station, radar, satellite, and other observational data, and on geospatial alignment to model grids.
- Build and operate a cycling data assimilation pipeline for our operational forecasting models, and produce the high-resolution gridded products it enables downstream.
- Develop observation quality control, bias correction (VarBC), and thinning workflows that hold up at operational data volumes and degrade gracefully when feeds drop out.
- Choose, deploy, and adapt a modern DA framework (e.g. JEDI/UFO, GSI, DART, PDAF) for our regional and global needs.
- Run cycling DA–forecast loops end to end lateral boundary conditions, SSTs, soil states, and spin-up at convection-permitting (~1 km) resolution over Indian sub-regions.
- Stand up rigorous forecast verification across deterministic (RMSE, bias, spectra) and probabilistic (CRPS, BSS) metrics.
- Tailor weather prediction models to renewable-sector needs, particularly solar (GHI) and wind generation (100m winds).
- Assist in training AI-based weather prediction models.
- Work at the intersection of physics-based modeling and machine learning hybrid physics–ML systems, learned parameterizations, and emulators.

## WHO YOU ARE

## REQUIRED QUALIFICATIONS

- A master's or PhD in geophysical sciences, physics, applied mathematics, computer science, statistics, or a related field. A bachelor's degree with 3+ years of relevant research or operational experience is also acceptable.
- Hands-on work with limited-area or mesoscale models such as WRF, MPAS, or comparable systems including dynamical cores, physics parameterizations, and boundary-layer/convection schemes configuring and running them end to end (domains, lateral boundaries, physics suites, spin-up and stability), tuning parameterizations, diagnosing systematic biases, and verifying against observations or reanalysis.
- Experience running convection-resolving simulations at high spatial resolution (~1 km).
- Demonstrated depth in data assimilation, evidenced by operational work, model contributions, research projects, publications, or technical reports.
- Hands-on experience across the DA toolkit: observation operators and error specification; variational (3D-/4D-Var) or ensemble (EnKF, LETKF, EDA) methods; cycling workflows and innovation statistics; and assimilation of satellite, radar, radiosonde, or station observations.
- Familiarity with existing operational forecasting models (IFS, GFS, BharatFS).
- Experience contributing to or maintaining model code, maintaining data assimilation pipelines or holding responsibility in an operational or quasi-operational forecasting pipeline.
- Experience working with TB-scale, high-dimensional observational and modeling datasets (reanalysis, satellite, radar, weather-station, and sounding data) and the geospatial pipework (grids, reprojection, masks) around them.
- Hands-on experience with widely used reference datasets such as ERA5, MERRA-2, IMDAA, IMERG/GPM, and GOES/INSAT/Himawari.
- Practical experience on High Performance Computers (HPCs).
- Fluency in the modern geoscience Python stack: xarray, dask, zarr, netCDF.
- Experience building reproducible, production-grade pipelines.
- Excellent written and verbal communication, including the ability to explain technical work to both domain experts and cross-disciplinary collaborators.

## NICE TO HAVE

- Prior work on projects specific to Indian geography.
- Familiarity with coupled earth-system models.
- Experience with any of: ensemble and probabilistic forecasting, regional downscaling, or subseasonal-to-seasonal (S2S) prediction.
- Experience working with operational forecasting agencies (IMD, NCMRWF, ECMWF, NOAA, etc.).
- Familiarity with AI-based weather prediction models and data assimilation techniques.
- Comfort using agentic AI tools to accelerate development.
- Publications in respected atmospheric, oceanic, or climate science venues.

## WHAT YOU'LL GAIN

- Part of development of weather forecasting models deployed for real-time applications.
- Experience working on hard, open-ended problems at the intersection of AI and physical infrastructure.
- Exposure to how teams set priorities and push the frontier of AI weather prediction.
- Close collaboration with a deeply technical team.

## WHY THIS ROLE

This role sits at the frontier of the AI weather revolution, applying modern machine learning to earth system modeling. The next decade of progress in weather and climate prediction will be built by scientists who understand the physics and the data and have learned to wield generative AI. You will work in data-sparse regions where data is heterogeneous, ground truth is incomplete, and progress requires both technical depth and first-principles thinking.

Working hours The team is distributed across India and the US, so expect a few hours of evening overlap with US Pacific Time and IST on most workdays.

## About Pravāh

## Company Overview
- **One-liner**: Pravāh builds foundational AI intelligence for the electric grid, helping utilities forecast demand, model grid constraints, and reduce operational risk in real-time.
- **Entity Type**: Private (Seed Stage)
- **Headquarters**: San Francisco, California, United States (with offices in New Delhi, India)
- **Founded**: 2025
- **Founders**: Mohak Mangal (CEO) and Dhruv Suri (CTO)

## Core Business
- **Primary industry**: AI-powered Grid Intelligence / Clean Energy Software
- **Target customers**: Electric utilities (B2B/Enterprise) in India, Germany, and the United States
- **Mission or purpose statement**: To build the foundational intelligence for the electric grid and solve "a problem that impacts billions" — making electricity cleaner, more affordable, and more reliable as extreme weather, EVs, and rooftop solar strain the grid.

## Products & Services
- **AI Engine for Grid Management**: A machine learning platform that gives utilities real-time understanding of grid behavior under stress, using graph neural networks, reinforcement learning, computer vision, and deep learning-based forecasting.
- **Probabilistic Grid Simulations with Reinforcement Learning**: Tests thousands of possible grid futures to identify risk and failure modes before they occur.
- **Mapping Grid Infrastructure Using Computer Vision**: Uses satellite and street-level imagery to map grid assets and rooftop solar, revealing blind spots in the distribution network.
- **Deep Learning-based Forecasting**: Models electricity demand and distributed generation across time horizons, capturing volatility that legacy methods miss.

## Market Standing
- **Valuation/Market Cap**: Not publicly disclosed
- **Key Metrics**: Raised **$7M** (seed round); 13 employees; works with utilities serving tens of millions of consumers across three continents.
- **Notable Investors**: Khosla Ventures (Vinod Khosla), Pear VC (Mar Hershenson), and Conviction
- **Growth Signals**: Rapid headcount growth (+7.7% monthly on LinkedIn), traffic growth of +251.1% monthly to the website, deployed with utilities in India, Germany, and the US, and expanding across three continents.

## Competitive Advantages
- **Proprietary AI stack purpose-built for energy grids** — including PowerGNN, a topology-aware graph neural network designed specifically for electricity grids.
- **Full-stack grid intelligence** covering forecasting, constraint modelling, asset mapping, and probabilistic simulation in a single platform.
- **Real-world deployments** across three continents, including utilities serving tens of millions of consumers.
- **Strong founding team** with Stanford origins, backed by top-tier Silicon Valley VCs (Khosla Ventures and Pear VC).

## Strategic Focus
- Scaling deployments to more utilities globally, particularly in the US, India, and Europe.
- Deepening the AI/ML research edge — publishing in top venues (e.g., assessing global ML weather prediction models, graph neural networks for grids).
- Hiring across engineering, data science, and power systems roles to accelerate product development and customer implementation.

## Why Work Here
- **Mission-driven**: Founders describe this as "a problem that impacts billions" — working on making energy cleaner and more reliable in an era of climate stress.
- **Small, high-impact team**: Only 13 employees with a flat structure (43% are founders/founding team members) — meaning every hire has outsized ownership.
- **Research-forward culture**: The company publishes academic research and employs ML Weather Scientists, Staff ML Researchers, and Power Systems Engineers — a blend of hard science and engineering.
- **Flexible work**: Hybrid workspace; offices in San Francisco (HQ) and New Delhi. Stanford-founded with strong academic ties.
- **Strong investor backing**: Backed by Khosla Ventures and Pear VC, providing resources and credibility.
- **Notable perks**: The team wrote a detailed Notion document for candidates explaining "what we are building, what is at stake, and why we are so excited about this problem" — signaling a transparent, mission-oriented culture.

## Sources
1. [pravah.com](https://www.pravah.com/)
2. [pravah.com/our-team](https://www.pravah.com/our-team)
3. [linkedin.com/company/pravah-ai-energy](https://www.linkedin.com/company/pravah-ai-energy)
4. [jobs.ashbyhq.com/pravah](https://jobs.ashbyhq.com/pravah)
5. [builtin.com/company/prav-h](https://builtin.com/company/prav-h)

## Other roles at Pravāh

- [Associate](https://feeny.ai/job/associate-pravah-delhi-zqkwzk9pam9t) — Delhi, India
- [Senior Full Stack Software Engineer](https://feeny.ai/job/senior-full-stack-software-engineer-pravah-delhi-mr3aq2x6mxsw) — Delhi, India
- [Forward Deployed Engineer](https://feeny.ai/job/forward-deployed-engineer-pravah-delhi-rw9kapqfd9b0) — Delhi, India
- [Pitch us a Role](https://feeny.ai/job/pitch-us-a-role-pravah-san-francisco-tx6njwv7hf31) — San Francisco, CA