Machine Learning Engineer · Currently @ Upper Hand

I build intelligent systems that ship to production.

I'm Harshwardhan Patil - an ML engineer working across applied AI, data science, backend engineering, and MLOps. I turn messy data into models, and models into reliable systems that create real business value.

  • Applied ML
  • Data Science
  • Backend Engineering
  • Production ML / MLOps
  • LLMs & NLP
View projects Download Resume
Portrait of Harshwardhan Patil
01 - About

Engineering ML that survives the real world

I'm a Machine Learning Engineer at Upper Hand, where I design, ship, and own recommendation systems and analytics that move real revenue. I hold an M.S. in Data Science from Indiana University and bring 3+ years of data-engineering experience building and hardening production pipelines.

My work sits where applied ML, data science, and backend engineering meet: training and deploying models, turning ambiguous data into decisions, and building the services and pipelines that keep those models honest in production. I care about systems that don't just score well offline - they hold up, and create value, in the real world.

Lately I've been building LLM-powered tools - chat interfaces over structured data, semantic search, and retrieval - alongside classic ML for forecasting, classification, and recommendation.

02 - Skills

The stack I build with

Languages

  • Python
  • SQL
  • SQLX
  • Go
  • Java

Cloud & Infra

  • Terraform
  • GCP
  • Heroku
  • Docker
  • Kubernetes
  • Linux
  • Cloud Monitoring & Alerting

Data & Pipelines

  • PostgreSQL
  • Pandas
  • NumPy
  • Matplotlib
  • Metabase
  • Kafka

Frameworks

  • PyTorch
  • TensorFlow
  • Keras
  • Scikit-learn
  • Hugging Face
  • LangChain
  • Flask
  • Streamlit
  • Spring Boot
  • Gin
  • Fiber

Algorithms

  • Regression
  • Classification
  • Clustering
  • Boosting
  • Ensembles
  • Recommendation
  • Neural Networks
  • Transformers
  • BERT
  • LLMs
  • Semantic Search

MLOps

  • MLflow
  • Git
  • Conda

Backend & Architecture

  • REST API Design
  • GraphQL
  • Microservice Architecture
03 - Experience

Where I've shipped

  1. Machine Learning Engineer

    May 2024 - Present

    Upper Hand · Indianapolis, IN

    • Built and shipped an event-recommendation system reaching 63% accuracy in production.
    • Developed an LLM-powered chatbot that lets customers query their own data in natural language, opening a new revenue stream.
    • Generated $800 MRR by transforming QuickBooks P&L data into core financial KPIs (MRR, ARR, YoY/MoM revenue, expense, and net-profit trends).
    • Designed a client-location demographics & economics dashboard that created a competitive edge worth +$2,500 ARR.
    • Built a job-monitoring tool for ETL pipelines to improve reliability and recovery.

  2. Data Science Intern

    May 2023 - Jul 2023

    Faculty Assistance in Data Science Program, Indiana University · Bloomington, IN

    • Built a scraping & processing pipeline over 8 years of Mental Health Treatment Facility records, surfacing regional growth and service-diversification patterns.

  3. System Engineer

    Nov 2020 - Jul 2022

    Network Decision and Support Database (NDSD) Team, Infosys Ltd. · Bangalore, India

    • Maintained 50+ SQL-based ELT pipelines, processing terabytes of structured and semi-structured network data across 5 source systems with automated failure alerting.
    • Improved query performance by 55% through execution plan analysis, targeted index redesign, and schema normalization across Oracle databases, reducing average PL/SQL procedure runtime from 20 min to 9 min.
04 - Projects

Selected work

A sample of ML and AI projects - from LLM tooling to classic modeling.

LLM · Tooling

Chat with SQL Database

An AI chatbot that lets users query SQL databases in plain English - with an auto-corrective loop and semantic search to interpret intent.

  • LangChain
  • Streamlit
  • LLMs
  • SQL
ML · Classification

Loan Approval Prediction

Comparative study of SVM, K-NN, Random Forest, and Logistic Regression for loan approval — Random Forest led at 81% accuracy. Deployed on Heroku.

  • Scikit-learn
  • Random Forest
  • Heroku
NLP · Deep Learning

Named Entity Recognition with BERT

Fine-tuned BERT (with LSTM, dropout, and gradient clipping) for information extraction, reaching 87% accuracy on CoNLL-2003.

  • BERT
  • Hugging Face
  • PyTorch
NLP · Classification

Email Spam Detection

An NLP pipeline (bag-of-words, TF-IDF, n-grams) with Naive Bayes, Logistic Regression, and Random Forest - 97.4% accuracy across a 3,000-email corpus.

  • NLP
  • TF-IDF
  • Scikit-learn
ML · Forecasting

Walmart Sales Forecasting

A Random Forest + XGBoost ensemble forecasting weekly sales across 45 stores from historical and markdown signals.

  • XGBoost
  • Random Forest
  • Pandas
ML · Clustering

Customer Segmentation

K-means clustering for personalized marketing, with exploratory analysis and visual storytelling in R (ggplot2, plotly).

  • K-means
  • R
  • ggplot2

More on GitHub

05 - Education

Academic foundation

M.S., Data Science

2022 - 2024

Indiana University · Bloomington, IN

Coursework: Machine Learning · Deep Learning · Data Analysis · Cloud Computing

  • President & Director of PR, Data Science Club
  • Adobe Student Ambassador · Luddy Outstanding Service Award (2023 & 2024)
  • 3rd place, GT-IDEA Case Competition (2023)

B.S., Computer Science

2016 - 2020

Shivaji University · Kolhapur, India

  • Co-Founder, Code-Space (programming club)
  • Head, Student Training & Placement Committee
06 - Certifications

Credentials

07 - Contact

Let's aspire and create to inspire.

Open to Machine Learning Engineer, AI Engineer, Data Scientist, Applied Scientist, and MLOps roles. The fastest way to reach me is email.

Get in touch Resume