• Led development of a distributed agentic data ingestion platform on Ray, increasing throughput 6× and enabling fault-tolerant execution at scale.

• Architected and launched AWS Marketplace integrations, owning registration, metering, billing, and IaC across API Gateway, Ory, SQS, and Lambda.

• Designed the platform’s unified authentication and metering gateway using Kong and Ory Hydra, supporting secure multi-tenant workloads.

• Modernized the orchestration engine with DLQs and retry mechanisms, reducing unrecoverable jobs by ~80% and improving system reliability.

• Optimized Ray execution performance using Actors to eliminate bottlenecks and improve overall throughput.

• Developed core GraphQL APIs for Articul8 AI’s LLM platform, enhancing type-safe schemas and caching layers.

• Built Knowledge Graph APIs for entity linking and semantic retrieval, significantly improving internal developer usage.

• Implemented a robust notifications system, facilitating cross-team collaboration and debugging across various departments.

• Under review at EMNLP, researched on mechanistic interpertability of RAG models on LLMs and SLMs.

• Built a RAG pipeline with Langchain Agents and Pinecone for context generation to see changes in parametric memory use.

• Prompt Engineered with GPT-4 for synthetic data generation with quality assurance reducing manual efforts by

Utilized Machine Learning, Data Analysis, and NLP techniques to create training datasets, resulting in a 40% improvement in dataset quality and accuracy.

• Conducted proof-of-concept (POC) projects with PyTorch and TensorFlow, evaluating 7 new models, and presented results to data science team, facilitating informed decision-making and driving innovation.