About

Projects Experience

• Scout (Katalon)

  Sep, 2025 - Present

  A 'vibe-testing' platform specialized for automating web pages created by 'vibe coding' apps, such as Lovable and Replit. The product intent is to allow individuals with little to no coding experience to be able to ship their prototype products to production with less friction and worry.

  • In charge of designing and implementing structured reporting using an orchestrated Mastra workflows for automated testing results, streaming and persisting back through AG-UI and CopilotKit.

  • Collaborated to test and integrate AgentCore and related AWS Bedrock APIs for live-streamed remote browser automation.

  • Researched with the team to test, experiment, integrate Model Context Protocol (MCP) tools as well as architect context engineering flows to easily score and optimize context performance.

  • Handled prompt optimization and evaluations to improve AI-driven testing accuracy through incorporating the GEPA optimizer, a gradient-based evolutionary prompt optimization technique, to refine prompts for better performance in vibe-testing scenarios.

• Discovery Inloop

  May, 2025 - Present

  Delivered a virtual staffing consultancy product featuring editable knowledge artifacts, visual mapping, configurable AI assistants, and specialized consulting/facilitator workflows within seven-week timeframe.

  • Led end-to-end development, product design and architecture, and a production release of Discovery Inloop within 7 weeks, delivering AI knowledge systems, facilitator ecosystem, and specialized conversation features.

  • Built AI & brain functionality through agent discovery data pipelines to replicate "second brain" functions with visualizations, implementation of personalized n8n bot integrations, and long-context memory.

  • Designed and shipped facilitator platform with onboarding with contract signing through fingerprinting and action provenance, application/management dashboards, engagement workflows, facilitator notepads, and AI evaluation pipelines.

  • Improved platform reliability and developer experience through internal devcontainer support, optimized Kamal builds, upgraded CI/CD with rolling deploys, caching and audit logging, secure sharing and authentication flows.

  • Implemented conversation and integrations features with multi-modal message support, downloads/transcripts, Turbo/Hotwire real-time updates, Stripe plan-based billing, n8n/Graphlit integrations, comprehensive tests and operational runbooks.

• Screenz.ai

  Feb, 2025 - Mar, 20251 month

  Screenz.ai is an AI-powered platform that revolutionizes the hiring process by providing automated interview solutions using advanced AI technology.

  • Developed the entire backend for an AI interviewing agent, delivering the project in 2-3 weeks.

  • Integrated Next.js for the frontend framework, ElevenLabs' API for voice synthesis, and Neon DB postgres for database management.

  • Ensured seamless consolidation of technologies to create a robust, scalable interviewing platform.

  • Designed a sync engine to pull and deduplicate transcripts from ElevenLabs, enabling easy time-series queries using Retool.

  • Implemented prompt and context injection flow scaffold for future-proofing later implementations, integrating directly with ElevenLabs.

• Ascenda Travel Assistant

  Aug, 2024 - Nov, 20243 months

  Developed an AI-powered travel assistant for a leading financial services provider, strategically built on their existing Ruby on Rails infrastructure rather than introducing Python, leveraging AWS Bedrock and enhancing Ruby's AI capabilities.

  • Led the implementation of a dual-system architecture combining reactive query responses via a core assistant and proactive, context-aware suggestions (based on user location and history) to enhance user engagement.

  • Worked extensively with and contributed to the LangchainRB library, evolving it into an enterprise-grade solution suitable for the project's complex AI requirements within a Ruby environment.

  • Designed and implemented a caching strategy using Redis to optimize response times for both the main assistant and the suggestion system while maintaining contextual relevance.

  • Ensured seamless integration with the client's existing Ruby-based security, authentication, and core services, maintaining high levels of security and leveraging established infrastructure.

  • Demonstrated strong engineering judgment by forgoing their technical lead's adoption of Python in favor of adapting existing tools (Ruby) for AI tasks, prioritizing system reliability, security, and maintainability.

• Fornax AI

  May, 2024 - Sep, 20244 months

  Led the development of Fornax AI, an innovative AI-powered tool designed to revolutionize how startup founders create and refine investor pitch decks, targeted as a white-label product for venture capitalist and related firms.

  • Integrated expert knowledge from seasoned investors and industry professionals into the AI model, ensuring high-quality, actionable feedback for users.

  • Implemented features to provide specific, data-driven suggestions for optimizing pitch deck content and structure, tailored to investor preferences.

  • Led the revamp of the subscription system, with the implementation of a tiered pricing model that better aligned with user needs and usage patterns enterprise-level plans for accelerators and VC firms.

• Umbrella Product

  Dec, 2023 - May, 20244 months

  Umbrella Product is an all-in-one productivity app integrating task management, calendar, notes, and chat functionalities.

  • Architected and implemented scalable APIs for task management, calendar integration, note-taking, and real-time chat, ensuring efficient data flow and synchronization across multiple productivity tools.

  • Engineered robust data aggregation and pipelines, particularly for ESPN, to automate calendar scheduling and enhance other productivity tools, demonstrating proficiency in ETL processes and data integration.

  • Provided expert consultation on AI integration, focusing on agentic use-cases and chat functionalities, bridging the gap between traditional productivity tools and AI technologies.

• Hedge Foundation - Financial Data Aggregator

  Dec, 2024 - Mar, 20254 months

  Engineered an AI-powered data aggregation system in Elixir to scrape public web pages and fetch API data, consolidating financial information (specifically bank interest rates) and structuring it using Large Language Models.

  • Developed concurrent data fetching mechanisms using Elixir's Task.async_stream and Flow for efficient scraping of websites and retrieval of data from financial APIs.

  • Implemented data preparation and aggregation pipelines to combine content from diverse sources before processing by the AI model

  • Integrated with OpenAI's API via langchain.ex to extract specific, structured data (like interest rates across different terms) from unstructured text and tables based on a defined JSON schema.

  • Contributed directly to the langchain.ex open-source library, adding critical support for structured outputs (JSON schema) to facilitate reliable data extraction using OpenAI models.

  • Designed the system to handle multiple banks and their abbreviations, mapping extracted data to the correct financial entities and outputting a standardized JSON format.

• Dwarves - Internal AI Brainery

  Mar, 2025 - Present

  Developed Dwarves' internal knowledge base and brainery, focusing on achieving continual learning and evolving understanding using AI and Model Context Protocol (MCP) servers.

  • Engineered a robust architecture utilizing an append-only TimescaleDB hypertable and flexible JSONB payloads to ensure historical data integrity and adapt to evolving knowledge structures without schema migrations, directly addressing catastrophic forgetting challenges and in-context-learning.

  • Implemented a system leveraging Large Language Models (LLMs) for processing, synthesizing patterns, and generating insights, mapping raw data to structured information and synthesized knowledge based on the DIKW pyramid.

  • Designed and integrated TimescaleDB continuous aggregates to efficiently pre-compute trends and patterns from large time-series datasets, enabling the LLM to analyze information scalably and generate new concepts and knowledge.

• Dwarves - memo.d.foundation

  Jul, 2022 - Present

  memo.d.foundation is Dwarves Foundation's internal knowledge base and brainery, developed using Elixir, Python, DuckDB, and Next.js.

  • Engineered a fully local and private search engine using DuckDB-WASM, implementing hybrid search techniques that combine full-text and semantic search for enhanced accuracy and speed, with real-time embeddings using Transformers.js, enabling on-the-fly semantic analysis of search queries directly in the browser.

  • Optimized search performance through advanced indexing strategies, including full-text search (FTS) and Hierarchical Navigable Small World (HNSW) indexing, significantly improving retrieval speeds for large datasets.

  • Developed a high-performance markdown processing system using Elixir and Flow, enabling parallel processing of large volumes of files for efficient content extraction, transformation, and loading into the knowledge base.

Skills

Volunteer Work