David Bichara
From Research to Production
Most AI stays in the lab. I put it to work.
Proteome-scale drug screening. Fortune 500 data infrastructure. Products used by 150K+ people in 175 countries. Three years out of undergrad.
13,991xFaster screening: months of compute reduced to hours
150K+Users across 175+ countries
92%Prediction accuracy across species
My PerspectiveI believe we're at an inflection point. The tools to understand biology at the molecular level finally exist. Protein language models can screen entire proteomes in hours instead of months. Education can scale to anyone, anywhere. But most of this stays in the lab, or in a pitch deck.
I build the infrastructure that gets it into production, because the gap between "possible" and "deployed" is where the impact is. My background as an engineer gives me technical depth. My experience building and running companies gives me the product instinct to see the whole system, the user, and the business outcome at once. I think about what AI can become, not just what it does today, and I work to make that future real.
TrackRecord
From Fortune 500 infrastructure to biotech drug discovery. AI systems built for production, not just prototypes.
Enterprise AI Infrastructure
Architected data platforms processing billions of records daily at a Fortune 500 financial institution. Delivered measurable cost reductions and performance gains recognized by C-Suite Leadership. At this scale, even single-digit efficiency improvements translate to millions in annual savings.
Billions of records/day. Measurable cost reduction.AI for Drug Discovery
Co-founded Synthyra to accelerate drug discovery with protein language models. I built the full production stack and run live demos on partner molecules. Off-target protein interactions cause 30% of late-stage drug failures, each costing hundreds of millions. Our system catches these risks 14,000x faster.
14,000x faster. Months of compute in hours.AI Products That Scale
Built CompSciLib, an AI-powered education platform serving 150K+ users across 175+ countries. Outperformed GPT-4o on domain benchmarks and was accepted into MassChallenge. Grew entirely through organic adoption.
150K+ users across 175+ countriesFeaturedWork
Experience
I started by researching how AI could understand proteins. My career took me through building enterprise data platforms that process billions of records. I co-founded Synthyra to bring both disciplines togethe.
2025 – Present
Synthyra
Co-Founder & Platform LeadBuilt the full production stack for protein language model inference. Leading product strategy, running partner demos, coordinating a 4-6 person team, and handling all company operations.
2023 – Present
Fortune 500 Financial Institution
Software Engineer IIArchitecting enterprise data platforms processing billions of records daily. Delivered measurable performance gains and cost reductions recognized by C-Suite Leadership.
2023 – Present
University of Delaware, Gleghorn Lab
Research Fellow, AI & Protein ScienceApplied protein language models to therapeutic discovery. Research published on arXiv (2025). This work led directly to co-founding Synthyra.
2022 – 2023
MIT Teaching Systems Lab
Machine Learning Research AssociateBuilt supervised ML models achieving 77% precision in detecting confusion from audio data. Research published in peer-reviewed venue.
2021 – 2025
CompSciLib
Co-Founder & CTOBuilt and scaled an AI-powered education platform to 150K+ users across 175+ countries. Outperformed GPT-4o on domain benchmarks.
2019 – 2023
University of Delaware
BSc Biomedical Engineering, CS Minor3.97 GPA · Magna Cum Laude · BME Distinguished Junior Award · Tau Beta Pi
Research&Publications
Data Systems That Fuel AI
Patent pending. No USPTO number yet.
Diffusion Sequence Models for Enhanced Protein Representation and Generation
Synthyra
Open-Source Models
Synthyra publishes protein language models as open-source tools. 20+ models with 40K+ downloads.
Gauging Public Engagement with Nature of Science Concepts Through COVID-19 Tweets
7+ citations3rd most online engagement
Tic-Tac-Toe
Tic-tac-toe was one of my favorite games to play with my sister growing up. This version uses minimax search with alpha-beta pruning, the same class of algorithms behind chess engines. It evaluates every possible future game state to choose the optimal move.
Choose your symbol
Let'sWorkTogether
If you're solving a hard problem with AI, whether in healthcare, enterprise, or beyond, I want to hear about it.