Before I wrote software, I ran gels, set up PCRs, packaged lentivirus, kept mammalian cell cultures alive, and sat through 3 a.m. timepoints. I spent years at the bench first, and that order matters. I do not build AI and ML tools for genetic engineering as an outsider. I build them to solve problems I have personally hit doing the wet-lab work myself, then validate them back against real biology.
I am an AI-native builder. I architect systems end to end, from data pipelines and ML and quantum models to deployed web tools, and I use frontier coding agents such as Claude Code as my implementation layer. That lets me ship research-grade software on my own, at a pace and quality that would usually take a team.
- Molecular biology and virology: six years of hands-on bench research
- Genetic engineering: plasmid construction, cloning, viral vector engineering
- Vector systems: lentiviral vector design, packaging, transduction, titer optimization
- Cell culture: established cell lines and primary HSCs, with transfection and transduction workflows
- Gene editing: CRISPR knock-in and knock-out design, shRNA-mediated knockdown
- NGS-based integration-site analysis: oligo design for integration-site mapping on both Illumina and Oxford Nanopore (ONT) platforms, including lentiviral integration-site analysis and retargeting
- Grant and proposal writing: genetic engineering research proposals for institutional and funding-body review
- AI-native systems architecture: designing and shipping agentic tools solo, using LLM coding agents as the build layer rather than a shortcut around understanding
- Quantum computing for genomics: physics-informed quantum circuits applied to DNA biophysics and mutation modelling
- ML for molecular biology: GNNs (GraphSAGE and GAT), protein language model embeddings (ESM-2), conformal prediction
- Full-stack delivery: from data pipeline to deployed web tool (Streamlit, PyPI packages) on Google Cloud (Vertex AI), comfortable with MCP-based tool integration
- Broad LLM fluency: hands-on comparative experience across Claude, GPT, Gemini, and other frontier models day to day
| Project | What it is | Status |
|---|---|---|
| PEN-STACK | End-to-end open infrastructure for programmable genome writing: a learned locus atlas for safe and durable integration sites, a curated atlas of roughly 33K genome-writing enzyme systems, an inverse-design Write Planner, and an agentic RAG question-and-answer layer served through a live web app. One completed component (IS110-family recombinase design) hit 5 of 5 pre-registered predictions across 1,029 candidates. | pip install pen-stack, manuscript in preparation (target: Nature Methods) |
| DISCERN | Clinical decision-support engine to catch misdiagnosis across inherited bleeding and platelet disorders that mimic each other clinically, for example Glanzmann thrombasthenia versus LAD-III. It flags cases where the wrong diagnosis would mean the wrong treatment, not just the wrong label. | Implementation complete, validation in progress |
| bio-firewall | Genome-writing-native biosecurity middleware that supervises agentic design AI. A five-axis screen across cargo, locus, edit type, germline, and scale returns allow, flag-for-review, or refuse, each with cited evidence and a signed design passport. Defensive, open-data reference implementation. | Open source (Apache-2.0), benchmarked, manuscript in preparation |
| QuBiS-HiQ | Physics-informed quantum circuit that encodes SantaLucia nearest-neighbour DNA thermodynamic parameters directly into gate angles, extracting features through Pauli-Z expectation values for melting-temperature prediction. | Validated on IBM quantum hardware, tested with CI, manuscript in preparation (target: Nature Communications) |
I would rather hand a reviewer a plot with error bars and an honest "not yet" than a headline that does not survive scrutiny, and I hold my own quantum-computing work to exactly that standard. Claims earn their place through pre-registration, blind validation, and results that hold up when someone else runs them. Overclaiming quantum advantage is easy. Earning it is not.
- Email: ahmedaneesm@gmail.com
- Institute: Gene Therapy Laboratory, School of Biosciences and Technology, VIT Vellore