From Prompt Engineering to Context Engineering
Why Managing Context Is the Key to Smarter AI
Building agentic AI systems, multi-agent orchestration, and scalable backend platforms.
I'm Advait Darbare — a senior at Arizona State University majoring in Computer Science. I'm passionate about building intelligent, autonomous systems that merge AI reasoning with real-world impact.
Exploring the intersection of agentic AI, multi-agent orchestration, and scalable backend systems. I love designing systems that can reason, adapt, and automate complex workflows.
From developing ML components at Prophecy to researching data observability at Acceldata, I've grown from technical marketing into full-stack AI system design — combining creativity, engineering, and AI orchestration.
Performed in-depth analysis of research papers on data quality and machine learning model performance, deriving critical insights to support strategic planning.
Created internal documentation (reports, presentations, flowcharts, wireframes) supporting marketing efforts for Acceldata's data observability platform.
Compiled and analyzed data quality metrics to demonstrate the value proposition of data observability in enhancing ML model efficacy.
Engineered a PySpark ML pipeline for linear regression in an end-to-end ETL workflow, integrating vector assembly, train-test splitting, model fitting, and performance evaluation.
Developed a Word2Vec component using PySpark for the Prophecy Platform, enhancing NLP capabilities with tokenization and embeddings.
Built a data ingestion component for sourcing datasets from Hugging Face, streamlining ETL processes for ML pipelines.
Designed and executed 22 TPC-H benchmark ETL pipelines, showcasing efficient data extraction, transformation, and loading at scale.
Production-grade research platform with a LangGraph multi-agent supervisor, 10 institutional-style reports (Goldman, Morgan, Bridgewater, etc.), real-time Schwab/Alpaca data, and HITL portfolio trading.
Contract-first validation, anomaly detection, SLO enforcement, and HITL remediation. LangGraph pipelines, Next.js 15 dashboard, and MCP integration for AI-assisted data workflows.
A multi-agent platform that quantifies semiconductor supply chain risk and tariff exposure using knowledge graphs and agentic RAG.
Live streaming pipeline that flags anomalous Wikipedia edit patterns via a real-time dashboard using Kafka, Flink, and ML scoring.
Automated tool that collects daily check-ins, summarizes them via GPT-4, supports semantic search, and posts updates to Slack for efficient team coordination.
Why Managing Context Is the Key to Smarter AI
Exploring the key design patterns behind agentic AI systems
From RAG to Agentic RAG — where retrieval meets reasoning
A comprehensive comparison of AI tool integration approaches
Hallucinations, Reasoning, and Evaluation Challenges
I'm always open to discussing new opportunities, innovative projects, or just having a conversation about technology.