Software Engineer · Applied AI
I build AI applications, from model to interface.
M.S. Computer Science at Rutgers · Class of 2027
I'm a software engineer and Computer Science master's student at Rutgers, graduating in May 2027. At CodeBench, I work on a learning platform inside JupyterLab, building React interfaces, Python services, and an AI tutor that guides students through programming problems.
Previously, I worked on machine learning for security at Forenzy Networks, where I helped reduce false positives in web attack detection from 23% to 4%. I've also published research on facial liveness detection at IEEE CICT and won two hackathons. I like working across the stack: understanding the data, building the backend, and making the result usable in an interface.
I work across the React interface, Python extensions, FastAPI services, and database migrations for CodeBench, a learning platform inside JupyterLab. Its AI tutor uses notebook context and guided questions to help students work through code. I also work on storing conversations and tool calls for replay, grading, and instructor analytics.
I’m studying machine learning, artificial intelligence, and systems, and applying that work at CodeBench. GPA: 3.75 / 4.0. Expected graduation: May 2027.
I normalized logs in a web attack detection pipeline, reducing false positives from 23% to 4%. I combined five BERT classifiers to reach a 97.2% F1 score on a pipeline processing over 2 GB of HTTP telemetry a day, and built a monitoring dashboard that cut security triage time by 65%.
I curated more than 100,000 fraud records from SMS, email, and WhatsApp, built a text preprocessing pipeline, and integrated a TF-IDF classifier into an analyst portal to support manual review.
I graduated with a 9.15 / 10 CGPA and published research on facial liveness detection at IEEE CICT 2024.
Software and AI engineering roles. Graduating from Rutgers in May 2027.
Understand the problem.
Build and test a solution.
Put it into practice.
ideas. code. shipped.
An AI interviewer that uses your résumé and previous answers to ask relevant follow-up questions. I bring together hybrid retrieval, Gemini, and ElevenLabs for the conversation, with isolated Docker environments for coding rounds and a weighted rubric for feedback.
A tool for identifying staffing gaps and planning physician shifts around credential requirements. Recommendations use explicit rules that a scheduling team can inspect. Cerevia won the Sevaro track at Rutgers Health Hack 2026.
My published research on detecting photo and video replay attacks against facial authentication. The approach combines YOLOv8, Dlib landmarks, and a KNN classifier, reaching 98.5% accuracy on a combined dataset of custom samples, NUAA, and MSU MFSD.
A web attack detection system combining five BERT classifiers, with a 97.2% F1 score. The pipeline processes over 2 GB of daily telemetry through Nginx and ModSecurity. Its dashboard brings together attack patterns, confidence scores, and model explanations for security review.
I trained this neural network on 60,000 handwritten digits. Draw a number from 0 to 9 to see its prediction and how the network responds. The model runs in your browser; your drawing stays on your device.
Context → agent → tool → feedback
Find relevant context with hybrid search.
{
"session": "composur_voice_04","intent": "follow_up_probe","retrieved_chunks": 6,"rerank_score": 0.94,"voice_stream": "elevenlabs_turbo_v2","rubric_pass": true,}I'm interested in software and AI engineering roles, especially work on AI applications, backend systems, and developer tools. I graduate from Rutgers in May 2027 and am open to relocation. Reach out about a role, a collaboration, or a question about my work.