Mark Elliot Palmer II

Contact and headline

• Name: Elliot Palmer (full name Mark Elliot Palmer II)
• Degree: B.S. Computer Science, Data Science Emphasis, University of Mississippi
• Location: Atlanta, GA

Profile

Computer Science graduate (B.S., Data Science emphasis, May 2026) with hands on experience in applied machine learning and computer vision. Built end to end gesture recognition systems benchmarked across CPU and GPU platforms under a latency budget, and trained temporal CNN models on custom video datasets. More than 3 years producing RLHF training data for frontier LLMs. Seeking roles in ML engineering, AI, or computer vision.

Education

University of Mississippi — B.S. Computer Science, Data Science Emphasis (Aug 2022 to May 2026)

• Oxford, MS. GPA: 3.5. Engineering program member.
• Graduate coursework: Computer Vision and AI (A), Machine Learning (B).
• Undergraduate coursework: Operating Systems (A), Algorithms and Data Structures (A), Data Science (A), Organization of Programming Languages (A), Business Analytics Programming (A).
• Clubs: AI and Machine Learning Club, Finance Club, Sigma Chi.

Alpharetta High School and Georgia State University — Dual Enrollment Coursework (Aug 2018 to May 2022)

• Alpharetta, GA. GPA: 4.0. Honor roll 4 years.
• Awards: FBLA 1st in Region (Business Plan Competition), 7A Georgia State Track Champion, final cut walk on candidate for Ole Miss basketball (2 seasons).

Experience

Data Annotator — Data Annotation Tech (Mar 2023 to Present, Remote)

• Produce RLHF (Reinforcement Learning from Human Feedback) training data for frontier LLMs across math, coding, and computer science domains.
• Evaluate and critique model outputs in STEM tasks, generating high quality preference and correction data to improve model accuracy.
• Author training examples focused on numeric reasoning, algorithmic correctness, Logistic Regression, and Classification models.

Sole Developer — Discount MX (Aug 2024 to Present, Alpharetta, GA)

• Built a 50,000+ product database on Neon (PostgreSQL) with daily updating ingestion from distributor catalogs and API feeds.
• Designed and shipped a full stack e-commerce application (REST APIs, admin dashboard, consumer storefront) in collaboration with non technical stakeholders.

Projects

CatalisLabs — M&A Deal Sourcing Agent (Solo, 2026)

Python, a frontier LLM, Streamlit, SQLite, Gmail MCP, Git.

• Built an autonomous sourcing system as three scheduled frontier LLM agents with web search (sourcing, signals, and enrichment) that research, score, and keep current dental industry M&A prospects, deduplicating against an established company pipeline before writing through a single validated writer enforcing schema and enum checks on every edit.
• Designed a CRM dashboard (Streamlit) with a deal evaluation cockpit and a Kanban pipeline board, surfacing portfolio level quality and concentration metrics for review.
• Closed the outreach loop end to end: deterministic draft generation, human approval workflow, and Gmail MCP integration for draft creation, with git backed persistence so edits survive restarts on an ephemeral cloud deployment.

Engagement details, pipeline composition, and deal figures are confidential and are not published here. Happy to discuss the architecture and engineering decisions.

One Gesture Recognition (OGR) — B.S. Capstone (Solo, 2026)

Python, PyTorch, OpenCV, MediaPipe, TCNs.

• Designed a real time gesture recognition pipeline (pose extraction, temporal windowing, TCN classifier) and benchmarked the FPS by accuracy Pareto frontier across four platforms (CPU only, NVIDIA T4, NVIDIA L4, Apple M1), projecting onto edge targets with per target scaling rules.
• Trained TCN backbones on pooled NTU RGB+D and custom recorded video data; implemented per video holdout evaluation to replace naive window level splits and produce defensible accuracy numbers.
• Evaluated multiple pose estimation models (YOLOv26 Nano and Medium, MediaPipe Holistic, MoveNet) for latency and accuracy tradeoffs on constrained hardware.
• Findings: F1 of 0.97 to 0.99 cross take (seen subjects, new scenes) falling to 0.71 to 0.89 cross subject, isolating subject level rather than scene level overfitting; the larger backbone's advantage disappeared cross subject. MoveNet Lightning was the only backbone above 30 FPS on CPU (47 FPS on a Colab CPU) and was the deployment pick, trading the lowest cross subject F1 of the four for portability. The speed by generalization Pareto frontier is the result.

Resume Chatbot Website — Grounded RAG Assistant (Solo, Jun 2026 to Aug 2026)

Python, FastAPI, LangChain, FAISS, Groq, Docker, Google Cloud Run, JavaScript, three.js.

• Built a grounded retrieval augmented chatbot over a markdown corpus (FAISS, local BGE embeddings) running a two pass drafter and judge pipeline, so no unverified answer reaches the page; deployed as a hardened container on Cloud Run, scaled to zero, with rate limits bounding the model budget.
• Cut cold start from the 10 to 15 seconds originally recorded to 5.83 seconds by moving the index build off the startup path and baking the embedding model into the image, so the page no longer waits on the chatbot to render.
• Measured retrieval and answer quality with a checked in evaluation set rather than by inspection; two changes that looked obviously correct were caught and reverted by it.

Lycem Ledger — Group Project (2024 to 2025)

SwiftUI, Django, MongoDB, React.

• Built the iOS client in SwiftUI for an insurance company platform (Django and MongoDB back end, React web front end), delivered in Scrum sprints with a four person team, managing stakeholder demos, sprint retrospectives, and timeline tracking.

Skills

• Machine Learning: PyTorch, scikit-learn, CNNs, TCNs, model training, transfer learning.
• LLMs and Agents: LangChain, RAG, embeddings, FAISS, prompt engineering, RLHF data pipelines, model evaluation.
• Computer Vision: OpenCV, MediaPipe, pose estimation, image classification, data augmentation, edge deployment.
• Data and MLOps: pandas, NumPy, SQL, PostgreSQL (Neon), statistics, data visualization, Docker, FastAPI, REST APIs, Git, Google Cloud Run, Jupyter, Google Colab, Streamlit, PySpark.
• Web and Mobile: React, SwiftUI, Django, MongoDB, Stripe, HTML/CSS, PHP.
• Languages: Python, Java, C++, JavaScript, SQL, PySpark, Swift, HTML/CSS, PHP.
• CS Foundations: algorithms, data structures, time and space complexity, operating systems.