Work.
Production AI systems and commercial products I’ve built or led — spanning target discovery, clinical-trial intelligence, digital twins, regulatory automation, and portfolio decisions.
- LiveAI/MLFull StackKnowledge GraphCompresses hypothesis-to-insight cycles from months to weeks, varying by disease area
End-to-end AI stack combining a custom UI layer (Knowledge Graph, ChatAlphaMeld, Bioinformatics, Literature Analysis, AlphaClinMeld) with 12 ML models — including NER, RE, evidence scoring, digital-twin CT predictors and landscape assessment — running on PostgreSQL + Neo4j and self-hosted LLM APIs.
- Early accessAI/MLFull StackDrug DiscoveryHypothesis landscaping compressed from weeks to hours, with traceable evidence at every step
Guided drug-discovery hypothesis-generation platform built on the AlphaMeld Knowledge Graph. Supports four starting points — disease-first, target-first, indication expansion, and combination therapy — through a step-by-step workflow (project → config → exploration → agent scoring → ranking → hypotheses). A project-aware Discovery Copilot accompanies every step. Multi-tenant SaaS with SSO, RBAC, and audit logging.
- ResearchAI/MLNLPAutomation97.34% accuracy on 188-question ACM benchmark
Two agentic workflow engines fine-tuned on biomedical tasks; outperformed leading general-purpose frontier LLMs; delivered richer step-by-step rationales critical for complex clinical questions.
SequenceFlow® ResearchNexus & Clinical Navigator
LiveAI/MLFull StackClinical Trials6+ integrated modulesEnd-to-end AI stack combining Knowledge Graph, Clinical Trial Prediction, Literature Analysis, Patient Digital Twin, TPP, TOP with 12 ML models running on PostgreSQL + Neo4j and self-hosted LLM APIs.
AlphaCompass Portfolio Intelligence
LiveAI/MLData ScienceDrug DiscoveryStandardized indication-prioritization across 200 active programsClinical-trial–centric SaaS module on AlphaMeld® graph analytics; benchmarks pipeline assets and quantifies comparative risk. Adopted by two BD teams within six months, standardizing evidence-based indication-prioritization across 200 active programs.
Digital Twins for Systemic Sclerosis (EUSTAR)
ResearchAI/MLClinical TrialsDeep LearningIn-silico control arms over a 23,000+ patient, 200+ center systemic sclerosis registryOutcome-focused digital-twin framework built on the EUSTAR systemic sclerosis registry — longitudinal records for 23,000+ patients across 200+ centers. A PyTorch deep/generative stack (a mask- and time-aware GRU-D sequence forecaster plus a conditional sequence-VAE twin generator) produces prognostic scores and synthetic placebo trajectories for organ-specific and composite clinical outcomes. Designed as an ML-generated digital control arm — a PROCOVA/ANCOVA covariate that lifts trial power and cuts sample size — aligned with FDA digital-twin and EMA/ICH MIDD guidance.
Clinical-Trial Digital Twins
ResearchAI/MLClinical TrialsDeep LearningIn-silico placebo arms demonstrated for Phase II Type 2 Diabetes trialsConditional variational auto-encoder (CVAE) pipeline that synthesizes patient-level digital twins for Phase II Type 2 Diabetes trials. Generated matched placebo trajectories and predicted drug-response curves — proving feasibility of in-silico arms for protocol optimization and hypothesis testing.
Neurological Biomarker Intelligence Platform
EngagementAI/MLAutomationFull StackBuilt for Global pharmaceutical companyMulti-agent pipeline aggregating pricing, availability, and distributor intelligence for diagnostic kits across global marketsFull-stack platform that collects, validates, and structures biomarker market intelligence — pricing, availability, and distributors — for neurological diagnostic kits worldwide. A six-layer, event-driven multi-agent pipeline pairs deep web research with schema-enforced LLM synthesis; every data point carries source lineage, a credibility rating, and a stated-vs-extrapolated confidence flag. Vision-first PDF extraction, resumable checkpointing, and a canonical-run batch model keep large refreshes auditable. Built on Next.js, Drizzle/PostgreSQL, and Inngest for event-driven orchestration.
Rare-Disease Asset Screening (AlphaProspect Scout)
MVPAI/MLFull StackDrug DiscoveryBuilt for Specialty pharmaceutical companyNarrows a ~9,000-disease rare-disease universe to a ranked longlist and a provenance-backed asset shortlistExternal-innovation asset-screening dashboard on the AlphaMeld platform. A disease-first funnel narrows the ~9,000-disease rare-disease universe into a ranked longlist and a focused asset shortlist, with provenance on every score, gate, and number — deterministic and computed values kept visibly distinct from LLM-proposed-then-confirmed ones. A React 19 + TypeScript front end runs over a Hono API and PostgreSQL + pgvector for similarity search and entity resolution. Demonstrated on synthetic, illustrative data.
- LiveFull StackAI/MLFinTechTracks ~10,000 clinical-trial drugs and AI-predicts the catalysts that move small-cap biotech stocks
Investment research platform for small/mid-cap NASDAQ biotechs where a single trial readout can move the stock 50%+. Syncs clinical-trial data (ClinicalTrials.gov, Clinical Navigator) with live stock and financials (Alpha Vantage), then runs LLM catalyst discovery and streaming AI predictions for upcoming readouts, PDUFA dates, partnerships, and conference events. Includes bulk catalyst briefings and an interactive research chat.
- LiveFull StackRegulatoryAI/MLBuilt for Artixio Pvt LtdDrug & biologic regulatory intelligence spanning 12+ agencies — FDA, EMA, PMDA, NMPA, Health Canada, MHRA, Swissmedic, and more
Secure, role-based platform for pharma/biotech regulatory affairs teams to search, analyze, compare, and manage global drug and biologic approvals. Covers pharmaceuticals, biologics, biosimilars, and generics across 12+ regulators. Combines role-based dashboards, pathway and classification guides, dossier preparation support, a centralized guidelines library, Q&A with expert responses, and LLM-powered document analysis.
Secondary Research Automation
EngagementAutomationAI/MLNLPAgent-orchestrated multi-step research, delivered to dashboard or ExcelFully agent-orchestrated, multi-step LLM process. Input parameters, run a series of LLM processes to extract information from the web, compile, and store in a dashboard or export to Excel.
Booking System Platform
EngagementFull StackSaaSBespoke booking platform delivered end-to-end to client specEnd-to-end booking and scheduling platform — custom-built to client specification with real-time availability, notifications, and management dashboard.
Regulatory Intelligence for ArtiXio
LiveAI/MLRegulatoryAutomationAuto-drafts submission dossiers for first-in-class device filings; opened a new non-pharma revenue channelAI assistant that mines global regulatory updates and auto-drafts submission dossiers; established a new non-pharma revenue channel.
What I help with
AI & Data Science
- — GenAI & LLM-RAG Systems
- — Agentic AI Workflows
- — Predictive Modeling & Image Analysis
- — Knowledge Graphs & NLP
Scientific Computing & Infrastructure
- — Cloud & HPC Architecture
- — Data Pipeline Engineering
- — Full-Stack Platform Development
- — Digital Twin Modeling
Life Sciences & Drug Discovery
- — Target Validation & MOA Analysis
- — Clinical Trial Optimization
- — Regulatory Intelligence
- — Drug Repurposing & Portfolio Strategy
Leadership & Strategic Consulting
- — Team Building & Mentorship
- — Partner & Client Engagement
- — Product Vision & Agile Delivery
- — Speaking & Thought Leadership