I’m Mohd Atif, a data science graduate from IIT Madras who likes owning problems end to end. I build forecasting and AI systems, investigate markets and competitors, automate repetitive work, and turn ambiguous questions into something a team can actually use.
The common thread is not a title. It is how I work: understand the problem, find the evidence, build what is needed, and stay close to the outcome.
Forecasting, statistical testing, data pipelines and applied ML where the output has to be useful beyond a notebook.
Research, growth experiments, automation and direct delivery when the job is less about a fixed process and more about figuring it out.
My experience has moved between commercial analytics, founder-facing work, client delivery, AI evaluation and open-source contribution.
Worked directly with the founder on growth and customer-acquisition experiments during an early scaling push.
Delivered production websites and internal automation. One engagement included an n8n workflow that extracted event data from partner Instagram posts against a fixed schema, flagged uncertain fields and routed output through review before use.
Contributing to an open-source evaluation framework for physical AI. Added shell completion and environment diagnostics, with tests and changelog entries prepared for upstream review.
Training and evaluating a voice assistant at volume, with emphasis on rubrics and failure modes that aggregate scores can hide.
Owned forecasting and analytics for live operations where the output was a purchasing and pricing decision, not simply a dashboard. Worked across 100K+ inventory records, pricing tests and SQL reporting automation.
I’d rather show the system and the decision it supports than list every technology used to build it.
A research pipeline that collects competitor information, validates structured observations, preserves evidence and turns the run into an analysis/report. This is the project that best bridges my two target roles.
Day-ahead electricity price forecasting for the German and Luxembourg bidding zone, designed around chronological evaluation and data quality rather than a single model score.
A local-first assistant that turns plain-language tasks into scheduled browser agents, with a preview and explicit approval step before execution.
An authorization layer for AI agents that evaluates actions against organisation-scoped policies and can require human approval before consequential actions proceed.
Not every problem needs a model. Not every business problem needs code. I’m comfortable deciding which one it actually needs.
Python, SQL, pandas, statistics, forecasting, experiments, backtesting and data-quality checks.
Competitor research, market investigation, evidence trails and structured extraction from messy sources.
FastAPI, Flask, PostgreSQL, React/Vue, Docker, automation workflows and tested AI backends.
RAG, structured extraction, provider fallback, evaluation and explicit boundaries around uncertain outputs.
n8n, Playwright, scheduled agents and reporting automation that turn recurring manual work into a process.
Translate analysis and technical work into a clear recommendation, next step or tool someone can actually use.
I studied Data Science and Applications at IIT Madras. My work has gradually moved toward the overlap between analysis, AI systems, research and execution.
I enjoy the early part of a problem where the answer is not obvious yet: figuring out what matters, getting the evidence, building a first version and tightening it until it is useful.
I’m looking for roles where I can combine analytical depth with ownership. Data science, founder’s office, strategy & operations, or an early-stage team where the work crosses functions.