Kamil Arif.
Need an End-to-End CV pipeline? an Agentic LLM integration? custom ML architectures?
Experience
AT&T- Sr. Specialist MTS | Artemis Network Planning
— Present
Developed algorithms to optimize AT&T's L2/L3 Network Topology across the US, optimizing for cost without compromising on latency agreements.
AT&T- TDP Internship & Full-Time | IPSEA
— , —
Developed an algorithm to reorganize AT&T's IP Address Space more efficiently, generating $9M in estimated value. Also worked on an Anomaly Detection Model to identify failures in DNS Servers.
NJIT- Grader | CS 370 Intro to AI
—
Grader for a course centered around Data Mining and Machine Learning. Graded for a total of 228 students over the course of 3 semesters. Took an active role in helping students learn from mistakes and build a strong intuition for AI/ML Concepts.
CDx Diagnostics- Intern | Cancer Cell Classification
—
Trained a ResNet-50 CNN to differentiate cancerous cells from debris in dyed microscope slides. Was responsible for data collection, annotation, training, and testing.
Projects
ML Model Topology Research
—
Implemented & Replicated results of "Topology of Deep Neural Networks" and "Are All Layers Created Equal?" as part of a larger independent study on introducing more complex activations into ML models.
NJIT Schedule Builder & Course Analysis
—
Contributed to a student project aimed at building a better schedule builder site for NJIT. Implemented Data Ingestion pipelines and performed a variety of analyses on NJIT course curricula and structure. Reverse-engineered Ellucian Scribe Language.
Back2GPT
—
Built a LLM Chatbot to answer questions specific to the class material and related topics. Involved mining data from approved sources and organizing them to allow for reasonable & meaningful references. Used Markprompt for data processing and model access. Currently deployed as a Discord bot in the official class server.
Built a worldbuilding & mapping tool for star-wars-esque sci-fi galaxies, allowing users to generate and edit graphs representing transit routes across the generated world. Included resource placements with parameters for location and density.
ForensicXR
—
Worked on a VR educational tool that allows Forensic Science students to explore and investigate digitized recreations of Staged Crime Scenes. ForensicXR was developed in the NJIT MIXR Lab, first as a class project and then a summer research project under the NJIT Provost Undergraduate Summer Research Fellowship until August 2022.