Nico Bautista Hobin

Available for independent consulting Python · Claude Agent SDK · TypeSafe SDK · Iceberg · Arrow

• agent-office open-sourcing 2026.11: a runtime for agent organizations. A gateway, planner, and router run long-lived agents over one streamcast log, so the whole system is replayable and queryable, and each agent holds its own narrowly scoped credentials federated from OpenBao.
• streamcast github.com/nhobin219/streamcast: a JSON WebSocket pubsub framework. Messages are schema-validated and appended to a litelink log before subscribers see them, so streams double as queryable tables; at-least-once by default, with opt-in exactly-once and offset-based resume.
• litelink github.com/nhobin219/litelink: an embedded Iceberg storage engine for append-only data. Appends are durable on commit and land as sorted Parquet in an Iceberg table on local disk or S3, with no server, daemon, or catalog service.
• penca github.com/penca-io/penca: an experimental version-controlled lakebase in Rust (Flight SQL, DataFusion) with zero-copy branching, time travel, and row-level versioning on one columnar copy. Its ideas carried into litelink and streamcast.

New York, NY

Data Platform

• Designed and built from scratch, largely single-handedly, a high-availability, zero-downtime Python data service that serves mission-critical risk data at about 500 requests/s under a partial on-prem requirement.
• Optimized production trading and research analytic query workflows across cloud-native and on-prem technologies.
• Served as a top interviewer and intern mentor for EQR, and onboarded new engineers to Data Platform.

Alpha Strategies

• Member of the central liquidity team providing execution services to all of Citadel's fundamental equities businesses, working with quantitative researchers to implement and test ideas for optimizing the system's trading behavior.

New York, NY · stealth

• Co-founded and co-led a generative-AI startup (product, engineering, fundraising). Earned Y Combinator interviews, pre-seed investment from SBXi/77.xyz, and design-partnership negotiations with Crunchbase.
• Led MVP development on AWS (EKS, Docker, Pulumi IaC); owned deployments and cloud costs.
• Built natural-language APIs and RAG frameworks on GPT-4, LanceDB, and Qdrant (Go/Gin, FastAPI, Slack Bolt), with semantic query caching to keep inference costs down.

Remote

• Built features for Appian's patented distributed data store (ADS) in q/kdb+, Go, and Java to improve its scalability and flexibility.
• Improved query performance with read-optimized snapshots and background clean-up processes that cut disk usage.
• Extended the ADS query language (GraphQL-inspired) with computational functions, and engineered high-throughput ingestion paths for bulk writes.
• Built a Python performance-testing framework in GitLab CI to validate query speed and reliability; onboarded and mentored engineers and resolved production issues.

New York, NY

• Engineered a kdb+ real-time stack that consolidated 10+ data sources into 3 normalized real-time streams, powering the back end of Virtu Financial's Open Technology Platform.
• Designed and implemented the gateway architecture: the core query routing and execution logic.
• Optimized ETL pipelines from SQL Server into kdb+, and built the integration layers: JSON REST services, AMPS streaming interfaces, and legacy C++/Java interop.
• Led client onboarding and database consolidation for a major fintech client (ITG), and owned production operations for global deployments: releases, incident response, and ITRS real-time monitoring.

Spent a summer in Guadalajara researching the city's innovation ecosystem. Interviewed entrepreneurs, government officials, and venture capitalists, and compiled a report on the findings.