Moheet AI Engineer & Computational Drug Discovery Specialist London, UK
My work bridges the gap between Generative AI and Molecular Reality. I build autonomous in-silico pipelines that solve specific biological, regulatory, and clinical constraints. From replacing banned cosmetic ingredients to designing novel antipsychotics, I do not just run models; I engineer high-certainty IP generation systems using Deep Learning, Structural Biology, and Agentic Workflows.
Therapeutic & Clinical AI (The Pharma Engine)
-
Bio-Fusion Suite | Multi-Modal Endometriosis Diagnostics Objective: Overcoming the 7-10 year diagnostic delay in women's health. The Tech: A multi-modal fusion model integrating Proteomic Biomarkers (IL-6, CA-125, TNF-alpha) with Phenotypic Symptom Data. Impact: Achieved 99.65% synthetic validation accuracy, outperforming single-modality baselines by 3.65% and demonstrating the power of fusing blood and pain data for precision diagnostics.
-
TAAR1-ECL2 Engine | Schizophrenia Therapeutics Objective: Designing stable agonists for the TAAR1 receptor (Psychiatry). The Tech: An evolutionary scaffold-hopping algorithm targeting the difficult ECL2 Lid region of the receptor. Impact: Engineered a pipeline that improved geometric compliance from 0% to 74.4%, optimizing ligand residence time for next-generation antipsychotics.
-
Filorexant Rescue | Clinical Trial Simulation Objective: Rescuing failed drug assets (Merck's MK-6096) using AI patient stratification. The Tech: A computational simulation proving that Phenotype Filtering (targeting MDD patients with specific Insomnia comorbidity) could have salvaged the clinical trial. Impact: Demonstrates a viable Precision Medicine pathway to revive billion-dollar failed assets.
-
Project Halo | Structural Biology (AlphaFold 3) Objective: Discovery of high-affinity Super-Agonists for the VN1R1 receptor. The Tech: Leveraged AlphaFold 3 for state-of-the-art protein structure prediction combined with blind docking simulations (AutoDock Vina). Impact: Automated the screening of ligand interactions to isolate compounds that theoretically surpass endogenous agonist efficacy.
Computational Cosmetics & Toxicology (The Consumer Engine)
-
Project Adalia | Generative Regulatory Compliance Objective: Solving the EU REACH ban on Lilial (a key fragrance molecule). The Tech: An agentic pipeline using Scaffold Hopping (Adamantane/Decalin cores) to generate structural analogs. Impact: Identified candidates with Zero CMR (Carcinogenic, Mutagenic, Reprotoxic) alerts, retaining the olfactory profile while reducing theoretical synthesis cost by 64%.
-
TyroSniper V2 | Melanogenesis Inhibition Objective: De novo discovery of high-potency skin-tone modulators. The Tech: Transfer Learning pipeline trained on ChEMBL bioactivity data and fine-tuned on the COCONUT natural products database. Impact: Identified novel inhibitors with predicted sub-micromolar potency (<1000nM), optimizing for low systemic toxicity.
-
TRPV1-Puzzle | Sensitive Skin Modulation Objective: Decoupling pain relief from body temperature regulation. The Tech: Designed Non-Hyperthermic Negative Allosteric Modulators (NAMs) targeting the S2-S3 pocket (PDB: 9P6B) using NVIDIA BioNeMo. Impact: A solution for Sensitive Skin products that soothes neurogenic inflammation without thermal side effects.
-
MoreLife | Retinaldehyde Stability Pipeline Objective: Creating a stable, non-irritating alternative to Retinaldehyde. The Tech: In-silico screening with Human-Homology Filtering (hTyr/RAR-gamma) and metabolic stability prediction. Impact: Ensures candidates meet MoCRA (Modernization of Cosmetics Regulation Act) safety standards before they ever touch a lab bench.
Algorithmic Social Systems
- WINKLET | Spatiotemporal Privacy Protocol Objective: A Dark Forest theory approach to social matching. The Tech: A privacy-first algorithm that eliminates user profiles. A match is defined strictly as the intersection of two independent Geospatial & Temporal Signals within a defined radius. Impact: Validates mutual real-world presence without data exposure, facial recognition, or public swiping. Restoring serendipity through code.
Core Tech Stack AI & ML: Python, PyTorch, Graph Neural Networks (GNNs), Transfer Learning, Random Forest. Bio-Compute: AlphaFold 3, NVIDIA BioNeMo, ESMFold. Cheminformatics: RDKit, OpenBabel, AutoDock Vina, PyMOL. Regulatory: EU REACH, IFRA Standards, MoCRA, FDA Clinical Data Standards.