ML research engineer building retrieval, representation-learning, and evaluation systems for scientific discovery.
A CHEESE case study where 3D electrostatic similarity prioritized NS1, a tetrazolic neurosteroid candidate, before synthesis. NS1 later showed strong activity and advanced to in vivo validation in an osteoarthritis pain model.
Synthesis-on-demand libraries expose vast chemical spaces through reaction templates and compatible synthons, so the key question is which candidates to evaluate under a small oracle budget. SynthonBench introduces reaction-native, budgeted search tasks where optimizers propose valid reaction/synthon tuples and are scored with top-k recall, enrichment, regret, and Pareto metrics. Accepted to the…
Make-on-demand design spaces are increasingly defined by reactions and synthons rather than enumerated molecules, which makes search, audit, and visualization a grounding problem. This work presents SynthonOR for query-conditioned factorized retrieval and SynthonMap for route-aware sampling, embedding, and product-mass cartography. Accepted to the ICML 2026 GenBio Workshop.
Neural thickets suggest that many task experts may lie dense around pretrained weights, but density does not mean those experts are easy to discover. This audit tests whether nearby experts can be found through tiny learned low-rank routes under deployable selection and matched total query budgets, producing a mixed but useful probe of selector-sensitive neighborhoods. Accepted to the ICML 2026…
A lightweight Python package for reaction-grounded synthon retrieval that stays intentionally simple: generic RDKit fingerprints combined with a binary OR.
A Wittgensteinian essay on chain-of-thought, state over tokens, form of life, and why reasoning traces are better understood as public marks with an operative role than as private language or transparent inner speech.
A Deep MedChem webinar on exploring ultra-large combinatorial chemical spaces with CHEESE SynthonGPT.
Our team at Deep MedChem won the DYRK1B challenge at 9ADD 2026 with a CHEESE ShapeSim + LightGBM workflow. This short note explains what mattered and includes the final presentation.
Final presentation from the 9th Advanced in silico Drug Design workshop 2026, covering our winning workflow for the DYRK1B virtual screening challenge.
A compact synthon-conditioned transformer for navigating makeable chemical space, grounded in vendor enumerations rather than hallucinated SMILES.
Introducing a cellular-automata-native benchmark, leaderboard, and dataset suite that scores reasoning systems on metadata-rich ARC-AGI style episodes.
A quick note on adapting ISONET’s Gumbel–Sinkhorn edge alignment to molecular substructure search, with links to the source notebook and paper.
CHEESE is a web-based shape and electrostatic search engine. If you want to drop it into your web page, blog post, CMS or a private dashboard, an iframe embed is the quickest path—no additional scripts required.
Ultra-lean BitBIRCH rewrite that clusters hundred-million scale molecular libraries with on-the-fly compression, optional C++ accelerators, and a batteries-included CLI.
There are two distinct technologies to search in databases of molecules. This post explains both with intuitive LEGO illustrations and presents a practical guide on when to use each.
A recent paper claims an established virtual screening benchmark to be 'fundamentally compromised' due to 'egregious data leakage, rampant duplication, and pervasive analog redundancy', rendering it 'unfit for its intended purpose' and 'calling into question previous results based on its use'. The catch: the authors canonicalized to non-isomeric SMILES, effectively collapsing stereoisomers into…
An open-source React application that extracts all references from academic papers using DOI, converts them to 5000+ citation formats, and enables bulk downloads. Built with OpenCitations API and modern web technologies.
This is my billion-scale vector engine for 3D shape & electrostatic molecular similarity. It turns ligand-based screening into nearest-neighbour lookups—so you can *search, rank, and visualise* huge chemical spaces in seconds.
Web pages used to feel like the most boring part of programming—until I discovered React. In this post I reflect on my coding journey from the backend to the frontend.
A dockerized deployment of EPAM's Ketcher chemical structure editor with iframe communication API, React 19, and production-ready configuration for seamless integration into web applications.
Using on‑chain metadata for every token minted on Solana in February 2025, we show that fewer than 0.01 % of fresh meme‑coins meet even the loosest criteria for legitimacy.
In Frieren: Beyond Journey’s End, magic isn’t just a tool or a talent—it’s a reflection of character, circumstance, and culture history. From obsession and duty to chance and inheritance, the reasons characters become mages are as varied as the spells they wield. This piece collects dialogue excerpts from the anime’s first season to show how magic, for each mage, holds a different meaning—and…
A talk at ACS Spring 2025 in San Diego
Visualisation of chemical spaces using CHEESE embeddings.
Fingerprints have long been the standard for representing molecules, but neural network embeddings are opening doors to new possibilities. This post explores where these AI-driven representations excel, from better capturing molecular similarity with metrics to addressing modalities that traditional fingerprints often miss.
A novel approach to virtual screening using 3D shape and electrostatics in a vector space.
A novel approach to virtual screening using 3D shape and electrostatics in a vector space.
An app to predict ESP and RESP partial charges for small molecules.
I tried, and failed, to get ViTs to work on the ARC challenge.
A Glami webinar hosted by MITON on the topic of utilizing embeddings for drug discovery in billion-scale databases.
A ChemSpace Webinar on CHEESE embeddings
A Github Action to label issues based on their duration.
Grand Hotel Abyss, Stuart Jeffries' collective biography of the leading figures of the Frankfurt School, tells the story of critical theorists—mainly Jewish intellectuals of the early 20th century—who gathered around the Institute for Social Research, also nicknamed 'Café Marx'. The time has come to free the Frankfurt School from its detractors. To fully appreciate its seriousness in the irony of…
A review of Glass and Hall (2008), which argues that modern science is shifting from hypothesis falsification (Popper) towards data-driven model building. The paper suggests posing 'good questions' as a more constructive alternative, a view discussed here in its historical and contemporary context, including the reproducibility crisis and big data.
Winkler’s book The Inmate of Jean Genet traces the fate of a remarkable outcast—criminal, prostitute, and homosexual—hailed as “the last accursed poet” and “the Orpheus of the gutter.” Figures like Sartre, Cocteau, and Picasso even pleaded his case before the French president.
An essay on the phenomenology of neural networks and the implications of inceptionism, submitted in partial fulfillment of the high school leaving examination. This work was awarded in a school essay competition called Chytrá Palice
A richer sandbox post that demonstrates Markdown, MDX, citations, images, tables, footnotes, and interactive components.