Mudit Nautiyal
Data & ML Engineer | GenAI, Analytics
Mudit Nautiyal
Data & ML EngineerGenAIAnalytics
Building scalable data pipelines and high-performance AI systems that turn data into action.
Data and AI engineer working across analytics, data engineering, data science, machine learning, and applied LLM and agentic systems. I build data pipelines, glass-box ML tooling, and LLM agents, and ship them as open-source packages on PyPI and live applications.
OPEN TO
Technical Arsenal
Selected Work
Featured
- Open-source glass-box AutoML on PyPI. Profiles, cleans, trains, and tunes models, then writes a report that explains every decision.
- Runs on pure scikit-learn, stays deterministic, and uses no LLM and no network calls.
- Flags when a result should not be trusted, with a nested cross-validation optimism guard and honest minority-class reporting.
- Open-source data-quality engine that catches silent pipeline failures before stakeholders do.
- Profiles datasets and runs YAML-defined checks, then detects drift with z-scores against historical baselines.
- Routes Slack alerts and shows pipeline health on a live dashboard. Works on CSV, Parquet, or SQL.
- Personal memory engine. Ask a question in plain language and get a warm, cited answer instead of a list of matches.
- Embeds entries locally with sentence-transformers into Postgres pgvector; citations are attached server-side so the model cannot invent text.
- Python, FastAPI, and Groq (Llama 3.3 70B) on Supabase. The public demo runs on a synthetic journal for privacy.
- AI medical-billing advocate. Winner of the AI For Good Hackathon at the University at Buffalo.
- Extracts every line item and CPT and ICD code from a bill with OCR, then flags duplicates, upcoding, and mismatched codes.
- Generates a formal, ready-to-send dispute letter behind a HIPAA consent flow.
- Trading-intelligence platform that unifies live data from 9 global exchanges into one Streamlit app.
- Technical analysis (RSI, MACD, Bollinger), NLP news sentiment, and an ML signal engine with 30-minute learning cycles.
- Backtests strategies across 1,000+ trades with drawdown and risk-adjusted metrics.
- YOLOv8 quality-inspection system that detects Q-Block alignment markers on printed payout tickets through a multi-gate validation pipeline.
- Capstone built for Nissha Medical Technologies. Team lead.
- Calibrated inference thresholds to 99.91 percent accuracy on the production dataset.
More on GitHub
Experience
Recommendations
I would confidently recommend Mudit for any data engineering, data science, or ML engineering role. He has the technical foundation, the work ethic, and the problem-solving mindset to make an immediate impact.
Adithya Narayanan
Assistant Professor of Teaching, University at Buffalo
His ability to combine strategic thinking with hands-on execution makes him stand out as a valuable team member and a true asset to any organization.
Meena Malhotra
Associate General Manager at HCL
He consistently demonstrated exceptional technical expertise in Informatica MDM, ETL pipelines, and Data Engineering concepts, making him an invaluable asset to the MDM team.
Nitin Singh
Informatica and IDMC Developer at HCL
Let's Build
Something New.
Always open to discussing product design, engineering, or partnership opportunities.