Skip to content
View ka234388's full-sized avatar

Block or report ka234388

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
ka234388/README.md

Banner

Hey 👋 I'm Karthika Ramasamy

Senior Full Stack & AI Engineer • AI Research Assistant (CV)
M.S. Artificial Intelligence, specialized in Computer Vision @ UCF • Ex-Infoscian • Full Stack Developer
Open to Full Stack / AI-Integrated Full Stack / Computer Vision Engineering roles

Typing SVG

LinkedIn GitHub Email


👩‍💻 About Me

I'm a Senior Full Stack Engineer with 10+ years of experience designing, building, and running enterprise Java/Python applications end-to-end — from backend services and REST/SOAP APIs to responsive React/Angular front ends and cloud deployment.

Over the last year, I've extended that full-stack foundation into AI, GenAI, and Computer Vision, completing an M.S. in Artificial Intelligence (Computer Vision) at UCF and shipping production-style applications that combine traditional full-stack engineering with:

  • Computer Vision: object detection, pose estimation, tracking, and video analytics (YOLOv8, MediaPipe, OpenCV)
  • GenAI / LLMs / Agentic AI: RAG pipelines, multi-agent orchestration, tool-using agents, prompt design and evaluation (LangChain, LangGraph)
  • ML Integration into Full-Stack Apps: taking a trained model and wiring it into a real backend, API layer, and UI — not just a notebook
  • Production Engineering: Spring Boot/FastAPI services, REST APIs, CI/CD, Docker/Kubernetes, AWS/GCP, monitoring and reliability

I like building the whole thing — the model, the API around it, and the interface a real user touches.


💼 Experience

Senior Full Stack Developer & AI Integration — COMPETE BePlayFuel — Feb 2026 – Present Built a full-stack, AI-powered video-processing application: integrated a YOLOv8 pose-estimation + MLP classification pipeline into the backend, translating model output into real-time scoring and user-facing feedback across REST APIs and a React.js UI. Deep Learning · Machine Learning · Python · Java · React.js · OpenCV · PyTorch · YOLOv8 · Flask/FastAPI · GCP · Docker

University of Central Florida — Part-time — 10 mos

  • Graduate Teaching Assistant — Jan 2026 – May 2026
  • Graduate Research Assistant — Aug 2025 – Jan 2026 Engineered scalable batch-processing and data pipelines for an Urban Pedestrian Detection, Tracking & Behavior Analysis platform — 47M+ structured detection records across 286 georeferenced scenes. Deep Learning · Machine Learning · Python · SQL · OpenCV · GCP/HPC

Machine Learning Engineer (Full Stack AI) — N2 Services Inc — Remote — May 2025 – Aug 2025 Built a cloud-deployed, AI-powered Multimodal Visual Recognition System: integrated a MediaPipe pose-estimation pipeline with an MLP classifier into a Python backend, delivering real-time scoring at 30 FPS through a responsive UI. Deep Learning · Machine Learning · Python · Flask/FastAPI · MediaPipe · Streamlit · GCP

Technology Analyst (Java Full Stack Developer) — Infosys — Chennai, Tamil Nadu, India · Hybrid — Mar 2022 – Jul 2024 Designed and built enterprise application components Java, J2EE, Spring Boot, REST/SOAP APIs, plus CI/CD pipelines and AWS migration support. Databases · Git · Java · J2EE · Spring Boot · AWS

Java Developer — N2 Software Services Private Limited — Greater Chennai Area · Hybrid — Aug 2014 – Feb 2022 Delivered scalable Java/J2EE applications across the full SDLC using Spring, Hibernate, and microservice patterns. Databases · Git

Full project details and metrics are in my résumé — happy to walk through any of these.


🚀 What I Build

  • Full-stack applications with an AI/ML model wired into the backend, not bolted on after
  • Real-time computer vision pipelines (detection, tracking, pose estimation)
  • Agentic AI workflows and tool-using assistants
  • Retrieval-Augmented Generation (RAG) systems
  • Medical imaging and diagnostic support platforms
  • Scalable REST/microservice backends deployed on AWS/GCP with CI/CD

🔥 Featured Projects

🏃 Human Motion Analysis System

Problem: Give users real-time feedback on movement quality from video.

Impact: Full-stack app (React.js + Flask/FastAPI) with a YOLOv8 + MLP pipeline in the backend, producing scored, annotated video feedback in real time.

🩻 Multimodal Visual Recognition System

Problem: Real-time pose recognition and correctness scoring at 30 FPS.

Impact: MediaPipe + MLP classifier integrated into a Python backend, deployed on GCP with session analytics and live feedback.

🧠 Visual Agent with Explainable UI Actions

Problem: Automate desktop form-filling transparently.

Impact: Computer-use agent turning UI perception into structured JSON plans, executing clicks/typing via a perception–reasoning–action loop.

🏙️ Real-Time Pedestrian Analytics

Problem: Detect and track pedestrians across intersections.

Impact: YOLOv8 + DeepSORT pipeline across hundreds of scenes with trajectory-based analytics.

⚖️ Gender Bias Detection with VLMs

Problem: Measure and explain gender bias in multimodal reasoning models.

Impact: LoRA fine-tuned VLM + Chain-of-Thought pipeline classifying bias with human-readable rationales.

🤖 StorySprout – Multi-Agent GenAI Storytelling

Problem: Turn family memories into personalized children's storybooks.

Impact: Multi-agent platform (Safety, Narrative, FactCheck, Pedagogy, Quiz, Visual agents) producing safe, illustrated stories.

💬 WhatsApp Notification Router – Agentic RAG

Problem: Classify multimodal messages into notify / digest / mute.

Impact: Agent router combining OCR, Whisper transcription, hybrid retrieval, and confidence calibration.

🩺 Multi-Disease AI Diagnostic Platform

Problem: Unified, interpretable medical imaging pipeline for multiple diseases.

Impact: Diagnosis across five conditions with Grad-CAM heatmaps explaining predictions to clinicians.


🛠️ Tech Stack

Languages: Java, Python, JavaScript/TypeScript, SQL Full Stack: Spring Boot, Spring MVC, Hibernate, REST/SOAP, React.js, AngularJS, Flask, FastAPI AI / ML: Deep Learning, Machine Learning, GPT-4, Gemini, LangChain, LangGraph, RAG, Vector Search (FAISS, Pinecone) Computer Vision: YOLOv8, MediaPipe, OpenCV, PyTorch Cloud / DevOps: AWS, GCP, Docker, Kubernetes, Jenkins, CI/CD, Git Databases: Oracle, MySQL, PostgreSQL, MongoDB


📫 Connect

LinkedIn GitHub Email

Full stack at the core. AI at the edge. Building systems that see, reason, and act.

Pinned Loading

  1. Human-Action-Recognition-Finetuned-on-3-types-of-CLIPS-Model Human-Action-Recognition-Finetuned-on-3-types-of-CLIPS-Model Public

    I have fine-tune models—CLIP, and SGLIP—on the Human Action Recognition (HAR) dataset using the same train/test split with 10 epochs. where the model CLIP leverages large-scale vision–language pre…

    Jupyter Notebook

  2. Lightweight-Nuclei-Segmentation-using-MobileSAM-with-LoRA- Lightweight-Nuclei-Segmentation-using-MobileSAM-with-LoRA- Public

    Jupyter Notebook

  3. Multi-Disease-AI-Diagnostic-Platform Multi-Disease-AI-Diagnostic-Platform Public

    Jupyter Notebook

  4. Object-Detection-inference-on-Faster-R-CNN-DETR-Grounding-DINO Object-Detection-inference-on-Faster-R-CNN-DETR-Grounding-DINO Public

    Code is running object detection across the full COCO val2017 dataset (5000 images). And for each image performed I run inference through Faster R-CNN, DETR, Grounding-Dino and Dino model and colle…

    Jupyter Notebook

  5. People-and-Vehicle-Segmentation-using-CamVid-using-SAM2-and-YOLO-Model People-and-Vehicle-Segmentation-using-CamVid-using-SAM2-and-YOLO-Model Public

    BASELINE VS IMPROVED PIPELINE

    Jupyter Notebook

  6. YOLOv8_PedestrianDetection_with_UCSD_dataset YOLOv8_PedestrianDetection_with_UCSD_dataset Public

    Jupyter Notebook