Diego Said Anaya Mancilla

A single exome sequencing run yields tens of thousands of genetic variants, and clinicians and cancer researchers need to find the handful that matter, usually without bioinformatics training or command-line tools. I designed and built VCF/Plotein at the Cancer Genetics & Bioinformatics Lab at LIIGH-UNAM to close that gap: a web app where a user uploads a VCF file and explores each gene's variants as an interactive lollipop plot along the protein, overlaid with protein domains and clinical databases (ClinVar, COSMIC, dbSNP, gnomAD), and filterable by consequence, domain, and sample, so candidate variants can be prioritized visually.

I was the sole developer and a co-author of the resulting paper in Bioinformatics (Oxford University Press, 2019), written with researchers from UNAM and Cambridge. In 2026, I modernized it for how clinical labs need to handle patient data today: raw genomic data never leaves the browser, with multi-gigabyte VCFs streamed and parsed client-side in a Web Worker; the stack moved from Nuxt 2/Vue 2 to Vue 3, Vite, Pinia, and D3 v7; and the UI was redesigned into a single clinical workspace.

I built this to show that a public blockchain can solve a concrete, everyday problem: sharing private files without trusting the server that stores them, with no smart contracts, tokens, or speculation involved.

Starting from a public tutorial, I added client-side encryption, a CLI, and an on-chain index: files are encrypted with AES-256 on the user's device, so the storage layer (IPFS) only ever sees ciphertext; the content ID that IPFS returns doubles as an integrity proof; and a zero-value Algorand transaction records it in an immutable, per-account file index for a fraction of a cent. Algorand featured the project on its Developer Portal as a reference solution for developers learning to combine blockchain with decentralized storage, and it is my most-starred public repository. In 2026, I brought it back after its dependencies were sunset (PureStake API shutdown, algosdk v1 to v2, the deprecated ipfs package replaced with a around 100-line Kubo RPC client), because a reference example that no longer runs stops teaching anyone. The live demo is a block-explorer view of the files indexed by the CLI.

As an undergraduate in 2015–2016, I saw how hard it was for Mexican high school students to choose a university and degree: official SEP (Ministry of Education) data was scattered across government sites and nearly impossible to search or compare.

I built Universidades México as a vocational-guidance tool to fix that: a Django app with a scraper that consolidated SEP records into one database, natural-language search (NLTK) so students could find programs by what they wanted to study, and a map to explore options by state, deployed on DigitalOcean. A server that needs patching and a monthly bill is hard to sustain for a free public service, so in 2026, I rebuilt it to run at zero operating cost: a static Nuxt 3 site on Cloudflare's edge covering 3,467 universities and 27,798 programs. Every university and degree has its own indexable page (approximately 22,000 in total), so a student searching Google lands directly on the program they care about; search is instant and typo-tolerant with no server round-trips; and the map shrank from 26 MB to 73 KB, so it loads well on mobile connections. The build normalizes the SEP data into around 7,500 canonical degree pages, with CI/CD on GitHub Actions and Vitest.