Junior Agentic AI Engineer and Computer Science student with expertise in Python, Large Language Models (LLMs), and autonomous agent frameworks such as LangChain and CrewAI. Developed multi-agent systems and agent architectures incorporating memory, planning, and tool-use functionalities for complex task automation. Implemented reinforcement learning algorithms, specifically Proximal Policy Optimization (PPO) and reward modeling, within Unity simulation environments to refine agent behaviors. Contributed to building Retrieval-Augmented Generation (RAG) components and fine-tuning models using PEFT and LoRA to improve reasoning and performance. Collaborated on the deployment of LLM services using FastAPI and Docker while creating scenario generators to test AI agents against diverse edge cases.
Junior Agentic AI Engineer and Computer Science student with expertise in Python, Large Language Models (LLMs), and autonomous agent frameworks such as LangChain and CrewAI. Developed multi-agent systems and agent architectures incorporating memory, planning, and tool-use functionalities for complex task automation. Implemented reinforcement learning algorithms, specifically Proximal Policy Optimization (PPO) and reward modeling, within Unity simulation environments to refine agent behaviors. Contributed to building Retrieval-Augmented Generation (RAG) components and fine-tuning models using PEFT and LoRA to improve reasoning and performance. Collaborated on the deployment of LLM services using FastAPI and Docker while creating scenario generators to test AI agents against diverse edge cases.
• Built an autonomous agent system using Python and LangChain to automate complex workflows and multi-agent task execution. • Implemented reasoning-based agents with autonomous tool usage and long-term memory via Pinecone to improve AI service capabilities. • Developed RESTful Application Programming Interface (API) integrations to enable communication between AI agents and external data workloads. • Evaluated model performance across various Large Language Model (LLM) providers to ensure the reliability of AI platform components. • Experimented with agent architectures involving memory and multi-step reasoning to improve task completion rates for data processing. • Tested prompt engineering techniques like Chain-of-Thought (CoT) to refine agent behaviors and ensure high-quality AI service outputs.
• Developed a real-time object detection system using Python and YOLOv8 to identify and track objects in dynamic data environments. • Applied computer vision techniques to process and classify high-precision data from diverse video feeds for tactical analysis. • Simulated AI-driven strategies within controlled testing environments to assess model performance and decision-making logic. • Designed a modular architecture for vision-based services, facilitating easier integration with event-driven architectures. • Created synthetic data generation pipelines to train models on edge cases and failure modes identified during observability testing. • Contributed to the development of scenario generators that produce realistic test cases for the evaluation of AI platform services.
• Developed an AI-driven drone simulator in Unity using ML-Agents and Proximal Policy Optimization (PPO) reinforcement learning. • Implemented autonomous behaviors through reward modeling and reinforcement learning (RL) training for navigation and aiming. • Designed stable physics-based flight simulations to improve training signals and reduce noise during serverless model optimization. • Integrated combat mechanics such as raycast weapons and health systems to create a robust environment for agent training and monitoring. • Used reinforcement learning algorithms to improve decision-making and fix model weaknesses identified during simulation runs. • Built feedback loops to identify failure cases in agent behavior and applied reward modeling to enhance overall model performance and reliability.
• Developed a computer vision application using Python to analyze food safety by evaluating complex ingredient lists on packaging. • Implemented YOLOv8 and EasyOCR for high-precision extraction and semantic parsing of text from images, supporting document intelligence goals. • Integrated a locally hosted Mistral Large Language Model (LLM) to perform reasoning and identify specific additives or ingredients. • Designed the system to function as a task-specific agent, utilizing reasoning-based logic to process data and generate safety reports. • Delivered a responsive user interface using Streamlit to provide a seamless experience for mobile and desktop users. • Optimized model inference performance to ensure efficient task execution and rapid output generation in production-like environments.