About Me

  • Full Name:Pratham Saraf
  • Phone:+1 347-793-7420
  • Email:prathamssaraf@gmail.com
  • Location:Brooklyn, New York

Hello There!

Hi there! I'm Pratham, a Computer Science graduate from New York University (MS, 2026) with a CGPA of 3.92/4. During my time at NYU I served as a Teaching Assistant for Deep Learning, Tutor for Numerical Computing at Courant, and Research Assistant in AI & Neuroscience. I also hold a B.Tech in Data Science Engineering from Manipal University Jaipur, where I graduated with distinction (8.91/10 CGPA) and received the Dean's List Excellence in Academics Award.

My work has been published at ACM SIGGRAPH 2026 — a technical paper on dichoptic foveation for VR rendering optimization. I'm also a two-time hackathon winner, including 1st place at the NVIDIA × Antler × Acer NYC Spark Hack, where my team built a real-time public-safety vision system deployed entirely on local hardware.

My research spanned facial morphing detection pipelines using FISWG 19-feature frameworks with Prof. Banerjee, and vision models for behavioral neuroscience with Prof. Froemke. My work focused on developing lightweight AI architectures, including distillation of 72B models into efficient 7B architectures for real-world deployment.

With professional experience at Hewlett Packard Enterprise, ONGC, and OnFees, I've designed server management systems reducing crashes by 30%, built predictive maintenance models lowering downtime by 25%, and developed real-time analytics platforms. I'm proficient in Python, C++, Java, JavaScript, and Dart, with expertise in TensorFlow, PyTorch, HuggingFace, and cloud technologies including AWS, Docker, and Kubernetes.

My Resume

  • Work Experience

  • Software Engineering Intern

    Hewlett Packard Enterprise (HPE), Mumbai - July 2023 - July 2024
    • Designed server management system using Isolation Forest, reducing potential crashes and vulnerabilities by 30%.
    • Built alert manager with Prometheus, Loki, and Grafana for real-time anomaly detection across distributed servers.
    • Optimized MicroStrategy analytics dashboard, reducing query latency 10x and improving performance for thousands of daily users.
    • Developed YOLO-based document classification pipeline (95% accuracy), automating sensitive info detection & masking.
    • Collaborated with cross-functional teams to deploy anomaly detection pipeline across 100+ enterprise servers.

  • Project Intern

    Oil and Natural Gas Company (ONGC), Mumbai - June 2023 - July 2023
    • Built predictive maintenance model (Random Forest, XGBoost), lowering equipment downtime by 25%.
    • Partnered with electronics team to refine feature selection, improving analysis precision by 15%.
    • Developed real-time analytics platform for sensor data visualization, increasing decision-making speed by 8%.
    • Streamlined sensor data pipelines using optimized batch processing, reducing storage overhead and improving query efficiency by 12%.

  • Software Engineering Intern

    OnFees, Mumbai - January 2023 - February 2023
    • Integrated news feed and chat widget into mobile app, boosting user engagement by 20%.
    • Contributed to 10 feature updates in EdFly app with a 15-member team, improving release velocity by 25%.

  • Education

  • New York University

    Masters in Computer Science - May 2026

    CGPA: 3.92/4.0
    TA: Deep Learning — Assisted 120+ graduate students with coursework, assignments, and lab sessions.
    Tutor: Numerical Computing — Courant Institute of Mathematical Sciences.
    RA: AI & Neuroscience Research — Worked with Profs. Banerjee & Froemke on facial morphing detection pipelines and vision models for behavioral neuroscience.
    Coursework: Big Data, Artificial Intelligence, Deep Learning, Opensource Development, Information Visualization, Design & Analysis of Algorithms

    View Official Transcript

  • Manipal University Jaipur

    B.Tech in Data Science Engineering - July 2024

    CGPA: 8.91/10.0 - Received the Dean's List Excellence in Academics Award
    Coursework: Data Structures, OOPS, Computer Networks, OS, Machine Learning, NLP, Artificial Intelligence, Big Data Analytics

    View Official Transcript

Research Experience

  • Facial Morphing Detection (NYU – Forensics AI)

    Prof. S. Banerjee, NYU - July 2025 - Present
    • Built ML pipeline using FISWG 19-feature framework and CLIP-based NLP integration for forensic biometric analysis.
    • Active Research: Exploring distillation of Qwen2.5-72B models into lightweight 7B architectures for faster, resource-efficient detection.
    • Developing advanced facial morphing detection systems using PyTorch, Qwen 2.5, Molmo 2, and HuggingFace transformers.

  • Computational Neuroscience (NYU Langone)

    Prof. Robert Froemke, NYU Langone Health - July 2025 - Present
    • Automated social behaviour quantification from video using ML, improving accuracy vs. manual annotations.
    • Active Research: Developing a computer vision model to trace mice movement patterns from video data, enabling precise behavioural mapping in neural circuit studies.
    • Implementing TensorFlow and OpenCV-based systems for behavioral analysis in neuroscience research applications.

  • VR/AR Rendering Optimization (NYU Immersive Labs)

    Prof. Sun Qi, NYU Immersive Labs - July 2025 - Present
    • Designed and conducted 2AFC psychophysics studies in Unity with 50 participants, validating dichoptic foveation technique achieving 30-40% rendering compute reduction while maintaining perceptual equivalence.
    • Built production Unity application using C# implementing dichoptic rendering pipeline with multi-threaded architecture achieving real-time 60+ FPS performance across diverse VR hardware.
    • Created optimized algorithms based on 3D mathematics implementing Gaussian blur kernels, unsharp masking, and binocular fusion models to optimize visual effects while maintaining perceptual quality.

Publications

Dichoptic Foveation

ACM SIGGRAPH 2026  |  Technical Paper

Designed and validated a dichoptic foveation rendering technique via 2AFC psychophysics studies with 50 participants, achieving 30–40% reduction in rendering compute while maintaining perceptual equivalence. Built in Unity (C#) with a real-time dichoptic rendering pipeline running at 60+ FPS across diverse VR hardware.

My Projects

Intelligent DNS Security Platform

GuardNet

Enterprise-grade DNS filtering platform blocking malware, phishing, and ads with real-time threat intelligence. Go-based DNS resolver with caching, achieving <15ms response times. Node.js API gateway + React dashboard.

AI Nutrition & Health Assistant

WholeSight

Cross-platform mobile app with AI-powered food recognition (>90% accuracy) and multimodal logging (camera, barcode, voice, text). Firebase backend, Clean Architecture + BLoC pattern.

Collaborative AI Coding Workspace · YHack 2026

Parachute

Real-time multi-user cloud IDE where every developer brings their own AI agent. An Orchestrator AI prevents conflicts, manages file locks, and routes agents to non-overlapping work. Monaco Editor + Yjs CRDTs + xterm.js.

Predictive Surveillance HUD · NYC Spark Hack

POI Deck

Person of Interest DECK/01 — local-inference tactical HUD for NYC predictive surveillance. Processes live feeds from 900+ traffic cameras using NVIDIA NIM, RAPIDS, and Qwen VL entirely on-device.

Autonomous Job Application Agent

resumatic

Autonomous pipeline that scrapes job listings, scores fit, generates tailored 1-page LaTeX resumes, and emails top matches — all while you sleep. Built in Python with end-to-end automation.

Urban Data Analytics Dashboard

NYC Urban Rhythm

Comprehensive dashboard combining NYC public datasets — weather, taxi pickups, subway ridership, 311 requests, and events. Interactive visualizations reveal how urban factors correlate across boroughs over time.

Expense Tracking Application

Expensify

Flutter expense tracker with SQLite persistence, Provider state management, expense categorization, monthly analytics, and data visualization for personal finance management.

Extracurricular Activities

1st Place — NYC Spark Hack

NVIDIA × Antler × Acer  |  2025

Built Person of Interest — an end-to-end ML system analyzing live feeds from NYC's 900+ public traffic cameras using a vision-language model (Qwen3-VL-30B) for dangerous activity detection. Fine-tuned on a custom annotated fight-scenario dataset; fully local deployment on Acer Veriton GN100 (128GB VRAM). Won an NVIDIA GN100 workstation and was invited to present at the NVIDIA AI LinkedIn Live showcase.

1st Place — NYC Running Hackathon

Betaworks, New York  |  May 2025

Won $1,000 cash prize for the best project built with chat.dev — a platform for programming on-the-go via phone. Presented at Betaworks following a 5K run-hackathon format.

IEEE India Council Hackathon Mentor

IC Hack 2023

Mentored 10+ participants at the IEEE India Council Hackathon, providing guidance in app development and offering technical expertise in mobile application development and software engineering best practices.

Technical Skills

Programming Languages

Python, C++, Java, JavaScript, Dart, SQL

Databases

PostgreSQL, MySQL, SQLite, MongoDB

Frameworks

TensorFlow, PyTorch, HuggingFace, Flask, React, Flutter, Hadoop

Cloud & DevOps

AWS (EC2, S3, Lambda), Docker, Kubernetes, Git, CI/CD

AI/ML & Generative AI

Computer Vision (YOLO, OpenCV), NLP (Transformers, CLIP), LLM (Gemini API), Model Training & Deployment

Tools & Platforms

Firebase, MicroStrategy, Linux, VS Code, Prometheus, Grafana

Hire Me!

With a strong background in data science and a passion for using technology to solve complex problems, I am excited to join your team and make a significant impact on your next machine learning or deep learning project. Let's build innovative solutions together.

Contact Me Download Resume

Contact