Adilbek Bazarkulov, Software Engineer

PloverDB is a fast database that loads a large biomedical knowledge graph (about 7 million biological concepts and 30 million relationships among drugs, diseases, and genes) into computer memory and serves it as a web service for the NIH Biomedical Data Translator project. It was originally written by Amy K. Glen, and I joined the maintainer team in early 2026. Read more

NCATS Translator Ingests is a shared data pipeline that pulls about 30 third-party biomedical datasets (drug, disease, gene, pathway, and literature sources) into the standard graph format used by the NIH Biomedical Data Translator project. Each source flows through the same download, transform, normalize, validate, and release stages, producing artifacts that downstream Translator services consume as their input. I joined the team in summer 2025 and own the ingests for the Gene Ontology Annotation database, SemMedDB, and PathBank, plus the upload and orchestration layer that ships builds to the rest of the project, and the companion web server that serves the released artifacts at kgx-storage.rtx.ai. Read more

RTX-KG2 is the build system that produces the biomedical knowledge graph (called KG2) used by the ARAX reasoning agent for the NIH Biomedical Data Translator project. The pipeline pulls about 20 source databases (drugs, genes, proteins, diseases, pathways, ontologies, literature), normalizes their identifiers, applies the project's shared category and predicate vocabulary, and produces a single canonicalized graph that downstream tools query through PloverDB. It was originally authored by a team led by Erica C. Wood and Amy K. Glen at Oregon State University; I joined the maintainer team in April 2025. Read more

ARAX is the modular reasoning agent that answers biomedical questions for the NIH Biomedical Data Translator project. Given a structured question about how drugs, diseases, genes, and proteins relate, ARAX dispatches the query to the right combination of about 40 knowledge providers, merges the answers, scores them, and returns a ranked result with the supporting evidence. It was originally authored by a team led by Amy K. Glen at Oregon State University and David Koslicki at Penn State; I joined the maintainer team in early 2025. Read more

Gestro is a browser-based paint studio where you draw with your hands in front of a webcam and an AI model refines what you drew on demand, repeatedly, in a loop. 1st place Innovator Track and 3rd place Best Project Overall at HackUP 2026 (University of Portland). Devpost I built it solo over the hackathon weekend and have been iterating since. Read more

SuppFindr is a same-category supplement comparison tool covering more than 20,000 products. The product takes a deliberately anti-leaderboard stance ("compare, don't crown") and surfaces three dimensions side by side (price per active ingredient, potency per serving, and pills per serving) instead of collapsing them into a single ranked score. A market-explorer view projects each product into a ternary plot of those three dimensions so the user can see tradeoffs at a glance, and the medical-disclaimer copy is explicit: this is an informational tool, not medical advice. I built and maintain it solo. Read more

PolyDebate is an AI-driven debate platform where multiple large language models (Claude, GPT-4, Gemini, and around 100 others through OpenRouter) argue the bull, bear, and adjacent positions on a live Polymarket prediction market in real time, each model producing its own probability estimate. Each argument is narrated through ElevenLabs text-to-speech and streamed back to the browser as the models respond. 3rd place Polymarket Track at QuackHacks 2.0 (University of Oregon, November 2025). Devpost I built it with one teammate over the hackathon weekend; my half was the cross-tier integration that wires the Vercel-hosted Next.js front end to the Railway-hosted Flask backend, the deployment plumbing for both ends, the Postgres support, and a series of feature fixes on both sides of the wire (admin endpoints, mobile burger menu, audio-playback timing, dependency-CVE bumps). Read more

moleXa is a free, no-signup web app for exploring molecules in 3D in the browser. The frontend is a Three.js scene that renders any compound in PubChem with the right bond geometry and atom colors; a small companion API proxies PubChem's REST endpoints so the browser can fetch them despite CORS, caches the responses to absorb PubChem's per-IP rate limit (five requests per second), and adds plain-English explanations of the properties PubChem returns. I built and maintain both repos solo, starting in July 2025. Read more

Greeney is a transportation carbon-footprint tracker. Connect a Gmail account and the app reads delivery and flight receipts (Uber Eats, DoorDash, airline confirmations), looks up the relevant distances through Google Maps for ground orders and a Haversine calculation between airport coordinates for flights, and converts the result into kilograms of carbon dioxide using per-mode emission factors. Form-entered Uber and Lyft rides feed into the same pipeline, and the dashboard renders a 3D globe of the user's flight paths plus impact tiles (kilograms of CO₂, miles, tree-years to offset). 1st place Data-Driven Solutions Track at BeaverHacks Spring 2025 (Oregon State University, hosted by the Google Developer Group Club). Devpost I built it with three teammates over the 36-hour weekend; my half was the backend (Flask, the emissions calculator, the Google Maps and Gmail OAuth integrations). Read more