About
I like the part of the job where nobody has decided what to build yet. Most of my work starts as a vague problem — can someone who doesn't write SQL just ask this database a question? — and ends as something deployed that a real person uses.
That usually means owning the whole line: scoping what's actually needed, building the pipeline, standing up the infrastructure it runs on, and then explaining the result to people who don't care what a transformer is.
Core Principle
"A model isn't finished when it converges. It's finished when someone who didn't build it can run it, trust it, and get an answer under real constraints."
What Drives Me
End-to-end ownership · Ambiguous problems · Systems that run in someone else's environment · Shipping over polishing · Knowing when to cut scope · Signal over hype
Projects
Ask Your Database in English — Natural Language to SQL
What if someone who doesn't write SQL could just ask? I built a system that turns plain-English business questions into working SQL against multi-table databases it has never seen — schema linking, candidate generation, self-repair, and consensus voting. It runs on open-source models only, so it can be deployed inside a company's own environment with no data leaving it.
Thesis in progress — NJITGeomagnetic Storm Forecasting — Surya Foundation Model
Can we predict space weather 72 hours out? I fine-tuned a 366M-param Vision Transformer (pretrained on NASA solar imagery) using LoRA to forecast the magnetic field conditions utilities and satellite operators watch before a storm — at 3.5x lower training cost than full fine-tuning.
Flight Data Analytics at Scale — Hadoop on AWS
What happens when you throw 118.9 million flight records at a 6-node cluster? Built a distributed analytics pipeline on AWS EC2 with Hadoop MapReduce — achieved 100% data-local execution and a 2.73x speedup over baseline.
Peon — Memory for AI Coding Agents
AI coding agents forget everything between sessions. Peon gives them a persistent, local-first memory: it captures decisions as you work, consolidates them with an LLM, retrieves by hybrid search, and audits itself daily. Ships as an MCP server, so it drops into any client.
View on GitHubPersonal Portfolio — vineetvora.dev
Designed and built a distinctive dark-themed portfolio from scratch — featuring canvas particle networks, kinetic typography, magnetic hover effects, and a custom cursor. Pure HTML/CSS/JS, no frameworks.
View on GitHubFace-Based Attendance System — YOLO + FaceNet
Replaced manual roll calls with real-time face detection and recognition. YOLO spots faces, FaceNet matches identities, and the system auto-generates attendance sheets — zero human intervention.
View on GitHubCore Competencies
Publications
Preventing Wildfires in Energy Transmission by Automatic Power Line Defects Detection Using Machine Learning and AI
Presented at IEEE Conference, NJIT — December 10, 2025
Developed a deep learning framework using Mask R-CNN and DeepLabV3+ to detect power line defects and vegetation encroachment from UAV imagery in real-time — enabling utilities to shift from reactive repairs to proactive wildfire prevention.
Read on IEEE XploreEducation
New Jersey Institute of Technology
Master of Science in Artificial Intelligence · GPA 3.83
Advanced NLP, Computer Vision, Information Retrieval, ML for Data Science, Big Data
VIT — AP University
Bachelor of Technology in Computer Science
Certifications
Contact
Let's Build
Something Great
Building something interesting? I'd love to hear about it.