Dmitriy Vasilyuk

Here's what Littlebird itself helped me come up with after going over my last few months' data: - Develop Littlebird's core AI: agentic systems, RAG, LLM integration, and evaluation frameworks - Lead AI team initiatives and participate in technical architecture and hiring decisions - Contribute across the full stack, focusing primarily on backend development

- Created AI assistant that automates time-consuming web research using LLM agents & advanced RAG - DDG's web app (https://docdocgo.streamlit.app) was featured by Streamlit; API access is also available - "Heatseek" research feature: agent that helps find specific hard-to-find information by sifting through more and more sites and refining search queries, provides a concise summary of promising results with links to original sources - "Infinite" research feature: agent that searches for more and more sources on the desired topic, writes iteratively improving reports, builds up a knowledge base from retrieved sources for follow-up questions, allows the user to refine the topic, adjust the report type and manually set search queries - Allows to create and chat with knowledge bases from local documents (plain text, PDF, Word, HTML) - Enables users to manage and share their knowledge bases, offering customizable access levels - Built with Python, Typescript, Langchain, Chroma DB, OpenAI API, AWS, GCP, Next.js, Streamlit, FastAPI

- Designed architecture of AI app that comprises multiple agentic LLM workflows to perform research on potential clients for sales teams, allowing sales agents to quickly prepare effective, personalized ways to reach prospects. - Implemented app’s backend and UI features: chat, research using LinkedIn and 10-K SEC filings, free-form Internet research, processing results with LLM to generate insights on prospective clients’ market landscape and likely needs. - Collaborated with founder to brainstorm and iterate on overall product design as well as specific feature design and prompt engineering. - Oversaw junior developer, demonstrating and offering guidance on best practices for Python development, software design, UI design, version control, team collaboration on GitHub and elsewhere; helped achieve greatly improved code quality.

- Developed a data processing pipeline for a corporate security consulting firm, adhering to data privacy protocols and cutting down processing time by 70% per client's estimate (reference available) - The app's features include: survey response extraction from PDF, benchmark calculation, loading into PostgreSQL DB, generation of detailed spreadsheets with benchmarking and graphical representations of data, UI to search/edit/filter DB entries and control spreadsheet generation parameters

- Build smart contracts for gasless cross-chain transactions using Solidity, Hardhat, Foundry - Build API (using Next.js) to generate quotes and relay such transactions

- Build DocDocGo Carbon, Google Chat App to chat w/ Carbon's Confluence documentation (using GPT-3.5/4) - DocDocGo Core (public version) available on GitHub and as a live demo (see Featured)

- Build smart contracts and extensive test suite (Solidity/Hardhat) - Integrate frontend with smart contracts and IPFS (React) - Audit codebase for vulnerabilities, gas optimizations, other improvements - Write documentation for codebase

- Build Bet Czar, a full-stack dApp for managing and enforcing bets on the blockchain (Solidity, React, TypeScript, MUI, Ethers) - Build a perpetual futures trading bot on Solana (Python) - Build tools for time-series analysis, portfolio optimization, visual reports, (Python, NumPy, Pandas) - Analyze and invest in DeFi projects using above tools

- Physics and philosophy blog (https://www.reasonmethis.com) - Topics include philosophy of language, anthropic reasoning, multiverse, entropy, consciousness

- Beat mid & high stakes limit hold’em games with winrate 1.9BB/100 over last 1M hands - Built applications (Python, C++) to help play, analyze strategies, alert to weaknesses