Completed my M.S. in Computer Science at ASU
Graduated with a 4.0 GPA after earning top-5% marks in all 9 courses.
Irish Mehta | Machine Learning Engineer, Applied Scientist & Researcher
I'm Irish Mehta
A Machine Learning Engineer
ML engineer · applied scientist
About me
I build useful systems from difficult ideas.
I’m a machine learning engineer and applied scientist who enjoys taking a problem from its first experiment to a product people can rely on.
My work spans production ML for supply-chain and consumer products, computer-vision and remote-sensing research, and hands-on projects in NLP and generative AI. I recently completed my M.S. in Computer Science at Arizona State University.
Why I build
I have a soft spot for automation: finding repetitive, frustrating work and building something that automates the mundane.
Experience & education
A track record of applied ML.
Aug 2024 – May 2026
M.S. Computer Science
Arizona State UniversityComputer Vision • NLP • Generative AI • Robotics
- Graduated with a 4.0 GPA, top-5% marks in all nine courses, and an Engineering Graduate Scholarship.
- Ran 20K+ computer-vision experiments on ASU HPC infrastructure while developing Mars foundation models with the Kerner Lab and NASA.
- Built analytical models and dashboards at Decision Theater and researched efficient uncertainty estimation at the ACME Lab.
Oct 2023 – Jun 2024
Machine Learning Engineer
O9 SolutionsPython • Spark • Supply-Chain ML • Client Interaction
- Productionised hybrid clustering to optimise inventory for 10 Fortune 500 clients, cutting E2E runtime by 20%.
- Delivered ARIMA-based line-speed forecasts for AB InBev, raising accuracy by 10% & plant throughput 5%.
Jun 2021 – Oct 2023
Machine Learning Engineer
Head Digital Works · promoted from Associate Data ScientistReal-time ML • MLOps • AWS • XGBoost
- Promoted to MLE in 15 months and named Innovator of the Year after improving engagement 30% and reducing fraud 8%.
- Built an event-driven AWS model-deployment service that enabled ML marketing campaigns, increasing daily active users 10% and revenue 0.25%.
- Shipped real-time retention and 5.8M-user lifetime-value systems, increasing daily transactions 5% and customer retention 15%.
2017 – 2021
B.E. Electronics & Communication
BITS Pilani
Electronics • Signal & Image Processing
- Mentored 8 junior undergraduates across academic, co-curricular, and extracurricular pursuits, supporting their transition from school to college.
- Led a 40-member photography team, managing event coverage, inventory, and post-processing during the 2018-2019 academic year.
- Co-founded and led a 20-member acapella crew that won competitions across India and performed at campus events.
Projects
Showcasing some of my favorite work and personal projects
SudoHint
React • JavaScript • CSS • GitHub Pages • Sudoku • OCR • Computer Vision • Responsive UI
A client-side Sudoku learning app that manages 81-cell game state in real time and supports browser-only image import with grid detection, perspective correction, cell segmentation, and OCR review. It also includes validation, timers, responsive mobile UI, and auto- or manual-candidate modes without a backend.
Multi-ATS Job Intelligence Platform
Python • LLMs • Qwen • Google Sheets API • GitHub Pages • Job Matching • Information Extraction • Data Pipelines
A multi-ATS job intelligence platform that ingests postings across 8.6K+ companies and 7 ATS platforms, filters 3.2K+ relevant ML and data roles, and publishes a searchable dashboard for daily job discovery using LLM-based relevance scoring and structured metadata extraction.
AutoFrame
Python • Vertex AI • Gemini • Computer Vision • VLMs • Android • ESP8266 • Automation • Human-in-the-Loop Systems
AutoFrame is an autonomous photography assistant that captures overview images, proposes better camera poses with Gemini reasoning, and iterates with Android and ESP8266 hardware support to optimize portrait framing.
MOMO
Python • Foundation Models • Remote Sensing • Computer Vision • Vision Transformers • Model Merging • Mars Science • Representation Learning
MOMO is a multi-sensor foundation model for Mars remote sensing that merges representations from HiRISE, CTX, and THEMIS and supports downstream tasks across a wide range of Martian orbital resolutions.
Mars-Bench
Python • PyTorch • Deep Learning • Computer Vision • Image Classification • Semantic Segmentation • Object Detection • Model Finetuning • Hugging Face • Benchmarking • Research
Mars-Bench brings 20 standardized datasets for classification, segmentation, and detection of Martian features like craters, cones, boulders, and frost, aiming to make evaluation consistent and to catalyze Mars‑specific foundation models.
Adaptive Learning Platform
Multi-Armed Bandit • Adaptive Learning • Reinforcement Learning • Machine Learning
An adaptive learning system driven by a Multi-Armed Bandit controller to optimize question difficulty in real-time. The architecture uses a Streamlit frontend, a Groq API-based question generator, and JSON for persistent progress saving.
DishCovery
Python • Google Gemini AI • Snowflake Cortex API • GCP • Next.js • TypeScript • Natural Language Processing • Vector Search • Full-Stack Development
DishCovery is a food discovery app that analyzes restaurant menus and enables natural language query-based filtering of dishes and restaurants.
Silkot Silicones
Next.js • React • TypeScript • SEO Optimization • Vercel • Web Development • Responsive Design • Frontend Development
Designed and developed a responsive website for Silkot Silicones, a leading manufacturer of silicone products. The website features a clean, modern design with a focus on usability and SEO.
CounterEcho AI
NetworkX • Graph Theory • Social Network Analysis • Counter-narratives • Louvain Algorithm • Symbolic Reasoning
A vulnerability-based strategic counter-narrative system that analyzes social network data to generate targeted counter-narratives based on user vulnerability scores.
Earnings Call RAG
Python • FastAPI • LangChain • RAG • FAISS • Vector Databases • Embeddings • Financial AI • Document Analysis • Retrieval-Augmented Generation
Earnings Call RAG Bot is a Retrieval-Augmented Generation system designed for analyzing and querying financial documents (Currently supports earnings call transcripts)
Candidate Recommendation System
Python • LLMs • NLP • Resume Ranking • BERT • Embeddings • Vector Search • ReRanking
AI-powered resume ranking system that intelligently matches candidates to job requirements.
Prompt Engineering
Python • Large Language Models • LLMs • Generative AI • Prompt Engineering • OpenAI • GPT • Claude • Prompt Optimization • AI Best Practices
Best‑practice prompts and tutorials for large language models.
Machine Learning from Scratch
Python • NumPy • PyTorch • Deep Learning • Neural Networks • Machine Learning Algorithms • Linear Algebra • Gradient Descent • Backpropagation
From-Scratch implementations of classic machine‑learning and deep learning algorithms.
Auto Storage Categorization
Python • Natural Language Processing • NLP • Text Classification • Google Drive API • Document Processing • Automation • File Organization • Text Analysis
NLP‑powered bot that auto‑sorts Google Drive files into logical folders.
Movie Recommendation System
Python • Scikit-learn • Machine Learning • Recommender Systems • Collaborative Filtering • Content-Based Filtering • Hybrid Recommenders • Data Science
Hybrid recommender merging content clues, ratings, and composite ranking.
Predictive Modelling for Composites
Python • SciPy • Statistical Modeling • Regression Analysis • Machine Learning • Optimization • Data Science • Predictive Analytics
Regression models to optimize composite manufacturing parameters.
AI Engineering
End-to-end AI products, multimodal systems, and intelligent interfaces
Machine Learning
Models, benchmarks, and learning systems built around real technical problems
Data Science
Analytics and predictive systems that turn data into useful decisions
Signals of progress
What’s new.
A few recent milestones from my work across applied machine learning, research, and AI engineering
AutoFrame received a best course project award
A software-assisted hardware system for automated photography.
MOMO accepted at CVPR 2026
A multisensor foundation model for Mars remote sensing.
Earlier milestones
2025Received the Best Poster Award at the NeurIPS ML for Physical Sciences Workshop.
Submitted a CVPR 2026 paper under the guidance of Dr. Hannah Kerner.
Built DishCovery, a food discovery app, at Sunhacks 2025.
Received acceptance for a NeurIPS paper in the Datasets and Benchmarks track.
Completed a summer internship at ASU’s ACME Lab focused on efficient uncertainty estimation in deep learning.
Submitted a NeurIPS 2025 paper in the Datasets and Benchmarks track.
Completed the Spring 2025 semester with a 4.22 GPA and top-5% marks in two courses.
Started working at Decision Theater at ASU.
Started volunteering at the Kerner Lab at ASU.
Contact
Let’s talk.
Interested in build something useful?
I’m open to machine-learning, applied AI, and research-engineering opportunities. If you’re building something interesting—or have a process that should be automated—send me a note.
ihmehta@asu.edu