• Architected and implemented agentic workflows for simulation-code generation, designing document-to-code pipelines that translate structured and unstructured inputs into executable outputs through multi-step LLM orchestration.
• Built and evaluated code-generation agents using custom agentic architectures, benchmarking performance, reliability, and developer ergonomics against Claude Code SDK to guide tooling and workflow decisions.
• Developed MCP tools and reusable Claude skills to extend agent capabilities, improve task specialization, and enable structured human-in-the-loop and automated execution flows.
• Leveraged LangChain and custom orchestration patterns to coordinate planning, tool usage, context handling, and multi-stage reasoning across backend AI workflows.
• Orchestrated long-running, multi-step agent workflows with Temporal and Prefect for durable execution, retries, and observability across document-to-code pipelines.
• Built large-scale data processing pipelines with Apache Beam and integrated GraphRAG and LightRAG for graph-structured retrieval over technical documentation to ground code generation.
• Built production-oriented backend services with FastAPI to expose agent functionality through clean APIs, supporting integration, testing, and scalable deployment patterns.
• Developed frontend interfaces in React to enable interaction with agent workflows, document ingestion pipelines, and generated code outputs for faster internal iteration and usability testing.

Tech Stack: Python, LangChain, Claude Code SDK, MCP (Model Context Protocol), Claude Skills, RAG, knowledge graphs, vector DB, GraphRAG, LightRAG, HippoRAG, Temporal, Prefect, Apache Beam, FastAPI, React, GCP

• Architected and deployed a multi-agent LLM architecture for automated content generation, leveraging LangChain/ LangGraph to orchestrate planner → agent → aggregator workflows with web scraping, search, and enrichment pipelines.
• Designed and scaled Retrieval-Augmented Generation pipelines combining Pinecone vector databases with Neo4j knowledge graphs, applying GraphRAG and LightRAG for graph-structured retrieval to enable context-rich, real-time reasoning. Applied Matryoshka Representation Learning to optimize retrieval efficiency while reducing storage cost.
• Built automated content-generation pipelines producing SEO-optimized articles on casino games, betting, and football for iGaming and betting platforms, tailoring tone and structure to each vertical.
• Designed and optimized casino game algorithms, balancing fairness, payout logic, and engagement mechanics while validating behavior through simulation and statistical analysis.
• Implemented custom inference pipelines tightly integrated with internal databases and deployed to AWS (EC2, Lambda, CI/CD) for low-latency, production-grade performance.
• Translated complex domain rules into machine-readable logic, creating structured representations for regulatory and policy-driven automation tasks. Built dynamic ETL pipelines with Apache Beam for real-time synchronization between PostgreSQL and Neo4j, orchestrated with Prefect and Temporal for durable, observable scheduling.
• Fine-tuned and optimized generative models using PEFT and LoRA techniques to achieve domain-specific outputs with minimal resource overhead.

Tech Stack: Python, LangChain, LangGraph, Pinecone, Neo4j, GraphRAG, LightRAG, Matryoshka Representation Learning, PEFT, LoRA, Apache Beam, PostgreSQL, Prefect, Temporal, AWS (EC2, Lambda, CI/CD)

Contributed to large-scale language model training focused on improving coding capabilities, reasoning quality, and production inference.

• Prepared large code-based training datasets supporting language model fine-tuning.
• Researched parameter-efficient transfer learning techniques that improved convergence while reducing computational requirements.
• Optimized inference performance and memory utilization for large conversational models.
• Implemented transformer-based improvements for code understanding and context-aware response generation.
• Evaluated model performance across diverse programming and natural language tasks to improve reliability and response quality.

Technologies: Python • PyTorch • Transformers • LLM Training