Hi, I'm Nikhil Soni
AI/ML Engineer | Data Scientist
I design and deploy intelligent, data-driven systems, combining deep learning, scalable backend architecture, and a commitment to clean code to solve complex, real-world problems.
I design and deploy intelligent, data-driven systems, combining deep learning, scalable backend architecture, and a commitment to clean code to solve complex, real-world problems.
I'm a Computer Science graduate from NYU, passionate about building scalable AI/ML pipelines and robust backend systems. I've interned at Emerson, Junglee Games, and HPE, and worked on impactful projects like CrisisCast and dementia prediction models.
My approach combines technical expertise with creative problem-solving to deliver high-quality solutions. I'm dedicated to writing clean, efficient code and creating intuitive user experiences that recruiters value.
Showcasing my technical depth in scalable AI/ML pipelines, vector databases, and full-stack integration.
Developed an end-to-end real-time pipeline that ingests social signals (Reddit/Google), classifies posts using a fine-tuned LLM, and visualizes crises on a dynamic dashboard. Leveraged Kafka, PySpark, FastAPI, MongoDB, and Qdrant.
Built a full-stack app fine-tuning Stable Diffusion XL with DreamBooth and LoRA for user-personalized AI images. Integrated a Flask backend and Gradio frontend for real-time image generation with a feedback slider.
ML-powered tool to estimate NYC rental prices using 300K+ listings. Developed a hybrid classification-regression model with 81% accuracy. Deployed via Flask and Dash with LLM-powered report generation for broker insights.
My technical toolkit: specializing in building and deploying production-ready AI/ML systems.
Professional roles and academic foundation that define my expertise.
Quant AI Research
Designed a production-grade AI pipeline that crawls markets daily, decodes financial sentiment with LLMs, and plugs directly into client-facing decision engines.
New York University (Deep Learning)
Mentored 400+ students through office hours, clarifying complex concepts in diffusion models, RL, and advanced deep learning.
Emerson
Built an LLM-based validation tool using T5 and One-Class SVMs to automate DeltaV code checks, saving significant manual QA hours.
Junglee Games
Extracted and modeled user gameplay data, deployed ML pipelines with AWS Lambda, S3, API Gateway, and tracked metrics.
Hewlett Packard Enterprise
Built a document management platform using Django, HTML, CSS, JS; integrated AWS services for storage and authentication.
New York University (Tandon School of Engineering)
Focus: AI/ML, Big Data, and Deep Learning. Coursework included: Deep Learning, Advanced Machine Learning, Data Mining.
Manipal University Jaipur
Graduated with Honors. Developed and published research on NLP and early dementia prediction.
