Published
IEEE ISTAS25
Accuracy
97% DNN
About Me
Building intelligent systems at the intersection of research and production.
I'm a Master's student in Computer Science at the University of the Pacific, specializing in Machine Learning, Generative AI, and AI systems engineering. I thrive at the boundary between research and real-world deployment.
From co-founding a GenAI trading startup (Stock Crusher) and launching an AI-powered iOS app (MealMuse) live on the App Store, to publishing research at IEEE ISTAS25 on fake account detection. I build things that work.
My research spans NLP, transformer models, autonomous robotics, and data engineering. I've worked as a Graduate Research Assistant across three labs simultaneously and as a Teaching Assistant for security analytics.
Based in the San Francisco Bay Area, I'm actively looking for roles in ML engineering, AI/software engineering, and research.
1+
Publications
IEEE ISTAS25
10+
Projects
Production-ready
97%
Accuracy
LIMFADD model
8+
APIs Integrated
In Stock Crusher
Education
M.S. Computer Science
University of the Pacific
Stockton, CA
Machine Learning, Generative AI, AI Systems
B.Tech Computer Science & Engineering
APJ Abdul Kalam Technological University
Kerala, India
Computer Vision, Deep Learning, Embedded Systems
San Francisco Bay Area, USA
Available for on-site, hybrid, and remote
Experience
From founding AI startups to graduate research and teaching. Here's where I've applied my skills.
Co-Founder & Software Engineer
Stock Crusher (GenAI Startup)
Summer 2025 – Present
- Built and deployed a production-ready multi-AI trading platform integrating 8+ real-time market APIs.
- Implemented a weighted consensus algorithm across Gemini, Perplexity, and OpenAI for BUY/SELL/HOLD signals.
- Delivered via Flask APIs and a React + Tailwind dashboard with resilient cloud deployment.
Co-Founder & Software Engineer
MealMuse (AI-Powered iOS Startup — Live on App Store)
Fall 2025 – Present
- Launched a production-ready AI-powered iOS app using SwiftUI and Node.js/TypeScript (Fastify) backend.
- Enabled personalized recipe generation via text, voice, and image inputs.
- Built scalable OpenAI GPT-4o-mini infrastructure with caching, reducing LLM calls by 80–90%.
Graduate Teaching Assistant
University of the Pacific (Advisor: Dr. Sethuraman Kuruvimalai)
Spring 2026 – Present
- Assisted in ANLT-293B: Introduction to Security Analytics.
- Supported hands-on instruction in Linux, virtualization, Splunk SIEM, and SOC-based security investigations.
- Created quizzes, graded assignments, guided lab sessions, and prepared course materials.
Graduate Research Assistant
University of the Pacific (Advisor: Dr. Pramod Gupta)
Spring 2026 – Present
- Developed MoodMirror, an AI-powered student emotional wellness tracking system.
- Integrated GPT-4o-mini diary generation with a fine-tuned RoBERTa emotion classification model.
- Built with Next.js, FastAPI, and PostgreSQL with interactive analytics for mood trend detection.
Graduate Research Assistant
University of the Pacific (Advisor: Dr. Solomon Berhe)
Summer 2025 – Present
- Conducted large-scale sentiment trajectory research on 324+ Reddit discussions within the ReleaseTrain.io ecosystem.
- Built a Python-based pipeline using VADER, REST APIs, and custom quality/reliability metrics.
- Compared author vs. community sentiment dynamics, validated through human labeling and confusion matrices.
Graduate Research Assistant
University of the Pacific (Advisor: Dr. Dongbin Lee)
Fall 2025
- Developed an AI-driven perception and control stack for the F1TENTH Autonomous Racing Car.
- Used Jetson Orin Nano and ROS 2 with computer vision and sensor fusion for real-time navigation.
- Implemented cross-architecture ARM64 Docker workflows for embedded GPU deployment.
Selected Projects
Production apps, research tools, and AI systems built from the ground up.
Stock Crusher
AI-powered multi-model trading platform aggregating real-time data from 8+ APIs with a triple AI consensus engine (Gemini, Perplexity, OpenAI) for BUY/SELL/HOLD signals.
MealMuse
AI-powered iOS app (live on App Store) generating personalized recipes via text, voice, and image inputs using GPT-4o-mini with 80–90% LLM cost reduction through caching.
LIMFADD
LLM-augmented multi-class Instagram fake account detection system classifying real, spam, scam, and bot accounts with 97% accuracy. Published at IEEE ISTAS25.
Real-Time NLP Sentiment & Toxicity System
Three-phase NLP pipeline analyzing 18K+ social media comments using transformer models (RoBERTa, BERTweet, BART) for sentiment, toxicity, and authenticity detection.
MoodMirror
AI-powered student emotional wellness tracking system using Next.js, FastAPI, and PostgreSQL with GPT-4o-mini diary generation and fine-tuned RoBERTa emotion classification.
Weather Data Engineering Pipeline
End-to-end multi-database weather analytics pipeline using MongoDB, ClickHouse, and Redis with incremental ETL, metadata lineage tracking, and real-time dashboard.
F1TENTH Autonomous Racing Car
AI-driven perception and control stack for autonomous racing using NVIDIA Jetson Orin Nano and ROS 2, with computer vision and sensor fusion for real-time navigation.
Crop Monitoring & Maturity Detection
Deep learning-based crop maturity detection system using a custom CNN model and drone imagery, achieving 86%+ accuracy in classifying plantation crop images.
On-Tree Areca Nut Fruit Maturity Detection
YOLOv5-based deep learning model integrated with DJI Mini SE drones to detect and classify areca nut maturity in complex backgrounds, optimizing harvest timing.
Pest Detection Using SVM
SVM-based pest detection system that preprocesses crop images and classifies infected vs. healthy tea leaves with high accuracy, enabling early pest identification.
Research
Published and ongoing research across NLP, fake account detection, sentiment analysis, and autonomous systems.
LIMFADD: LLM-Enabled Instagram Multi-Class Fake Account Detection Dataset
Introduced a novel multi-class Instagram fake account detection dataset using LLM-based augmentation. Classifies accounts into real, spam, scam, and bot categories with 97% DNN accuracy. Integrated LIME-based Explainable AI for model transparency. Outperforms existing Kaggle benchmarks.
Santa Clara University, CA | International Symposium on Technology and Society
Sentiment Trajectory Analysis in Software Update Reddit Discussions
Conducted large-scale sentiment trajectory research on 324+ Reddit discussions in the ReleaseTrain.io ecosystem. Built a Python pipeline using VADER to compare author vs. community sentiment dynamics over time, validated through human labeling and confusion matrices.
Dr. Solomon Berhe, University of the Pacific
Real-Time Social Media Sentiment & Toxicity Analysis
Three-phase NLP research pipeline achieving 90.23% sentiment accuracy on 18,214 YouTube comments. Implemented zero-shot toxicity classification into 7 categories using BART Large MNLI without fine-tuning. Supervised by Professor Tapadhir Das.
Prof. Tapadhir Das, University of the Pacific
AI-Driven Perception Stack for F1TENTH Autonomous Racing
Developing an AI perception and control stack for F1TENTH autonomous racing using NVIDIA Jetson Orin Nano and ROS 2. Implements computer vision and sensor fusion for real-time lane detection, obstacle avoidance, and object tracking on embedded GPU platforms.
Dr. Dongbin Lee, University of the Pacific
Skills & Technologies
A broad and deep technical toolkit across AI, full-stack development, robotics, and data engineering.
Machine Learning & AI
Programming Languages
Backend & Frameworks
Databases & Data Engineering
DevOps & Tools
Robotics & Embedded
Data & Visualization
Cybersecurity
APIs & Data Sources
Get In Touch
Open to full-time roles, research collaborations, and interesting conversations.
Let's Connect
Whether it's a job opportunity, a research project, or just a chat about AI. I'm always happy to connect. Based in the SF Bay Area and open to remote work.
manumathewjiss18@gmail.com
Phone
(209) 792-4475
linkedin.com/in/manu-mathew-jiss
GitHub
github.com/manumathewjiss
Google Scholar
Google Scholar Profile
@manu.mathew.jiss
Instagram (American Snaps)
@american.snaps
YouTube
@americansnaps
Send a Message
Fill out the form and it will open your email client pre-filled and ready to send.