Kemal Özkırşehirli / Projects

Projects

Research, engineering, writing, and experiments worth making public.

No projects match that filter.

Independent Geometric Deep Learning — Methodology-First 3D Mesh Learning - Jun 2026 – Present - Leading a 7-member international research collaboration developing a new methodology for 3D mesh and point-cloud learning. The project is methodology-first: literature synthesis, kNN graph pipelines, Point Transformer v3-style architectures, reproducible experiments, and manuscript planning come first; life-science applications are pursued if and only if they provide a principled fit for the method rather than a forced benchmark.

Geometric Deep Learning - 3D Meshes - Point Clouds - Methodology - Life Sciences

Code soon-to-be released - CV

PRISM AI Safety Fellow — Protein Foundation Model Red-Teaming - May 2026 – Present - Selected from 800+ applicants for PRISM 2026; developing adversarial evaluation methods and safety-bounded benchmarks for protein foundation models, including biological plausibility, sequence–structure–function constraints, developability, uncertainty calibration, claimed mechanisms, and failure-mode taxonomies.

AI Safety - Protein Design - Evaluation - Benchmarking

Code soon-to-be released - CV

ChromoGen-Engine V2 — Conditional Diffusion for 3D Chromatin - May 2026 – Present - Developing ChromoGen-Engine V2 with MIT Chemistry / CSAIL collaborators: a conditional diffusion and evaluation pipeline for small-molecule-perturbed single-cell 3D chromatin generation, integrating Plate-C / Easy Dip-C treatment structures with CHARM-style same-cell Hi-C, RNA-seq, ATAC-seq, histone-modification data, perturbation labels, genomic sequence, and regulatory-state features under leakage-aware splits.

Diffusion Models - Genomics - Chromatin - Single-Cell

Code soon-to-be released - CV

A*STAR V2M-Engine — Autoimmune Target Discovery - Apr 2026 – Present - Architecting a calibrated variant-to-mechanism target-discovery framework for autoimmune disease: GWAS/fine-mapped variants → regulatory programs → causal immune states → targetable disease mechanisms using single-cell foundation models, QTL/colocalization evidence, known therapeutic-target controls, perturbation validation, and structure/affinity-aware prioritization.

Drug Discovery - Single-Cell - Genetics - Target Discovery

Code soon-to-be released - CV

TBXT Small-Molecule Hackathon — Student Principal Investigator - Apr 2026 – Present - Leading an 11-person chordoma/TBXT computational hit-identification sprint: 2,274 prior-art compounds + 737 analogs + 30,000 BRICS recombinations → 570-compound novelty-filtered pool → 137 strict-pass → 24 submission-ready → 4 judge-facing site-F picks, with docking, QSAR, GNINA, Boltz-2, MMGBSA/FEP, selectivity, cost, PAINS, sourceability gates, and a public research-release repository.

CADD - Docking - QSAR - TBXT - Drug Discovery

GitHub - Group repo - CV

ChemAgent-QSM — LLM and Statistical Mechanics/Theoretical Chemistry - Jan 2026 – May 2026 - Developing ChemAgent-QSM, an LLM-guided agentic framework that turns natural-language chemistry prompts into validated quantum/statistical-mechanics Python workflows: PySCF electronic structure, optimized geometries, descriptors, vibrational spectra, time-correlation functions, local order metrics, relaxation timescales, mobility fields, and dynamical-heterogeneity analysis.

LLM Agents - Quantum Chemistry - Statistical Mechanics - PySCF

GitHub - CV

EVEdesign — Functional Protein Discovery and Algorithm Engineering - Apr 2026 – Present - Contributing algorithmic and engineering components to EVEdesign, an open-source method-independent protein-design platform across a 21-person, 18-institution, 8-country collaboration, emphasizing uncertainty-aware candidate selection, sequence–structure–function objective integration, protein-language-model/evolutionary-prior scoring, and lab-in-the-loop biosequence design.

Protein Design - Open Source - Uncertainty - Biosequence Design

Code soon-to-be released - CV

Pedal AI — LangGraph Agentic Chemistry Platform - Jul 2025 – Dec 2025 - Head of Chemistry Research-Implementation and backend software engineer: designed a LangGraph/FastAPI backend connecting chemistry agents to molecular prediction engines across 3 backend layers and 4,000+ lines of code; reduced prediction latency to <50ms, supported 15,000+ daily API requests, processed 50+ GB of chemical datasets, and built retrosynthesis UX with inspectable route trees, reaction provenance, and decision-ready synthesis workflows.

AI Agents - Backend - FastAPI - Retrosynthesis - Chemistry

Code soon-to-be released - CV

Deep Reinforcement Learning for Antibody-Antigen Interactions - Jan 2026 – May 2026 - Co-led a six-person AI@MIT/AIM Labs project building a structure-informed cross-attention transformer and ESM-2 pipeline mapping pathogen/protein sequences to candidate antibodies; curated OAS, SAbDab, IEDB, and CoV-AbDab-style datasets and reduced training time by ~40% with PPO against AF-Multimer/IGFold structural metrics on MIT SuperCloud/PyTorch DDP infrastructure.

Reinforcement Learning - Antibodies - Protein Models - PyTorch DDP

GitHub - CV - AI@MIT

ML for Statistical Mechanics Simulations — Prof. A. Nihat Berker - Jan 2022 – Present - Invited to research after ranking top 1% in graduate-level 8.334 while in high school; building a PyTorch/GNN-augmented Kadanoff renormalization group framework for spin-1/2 antiferromagnetic systems, automatic fixed-point discovery, and phase-boundary mapping; also mentored 200+ students through chemistry, mechanics, and career-path instruction.

Statistical Mechanics - GNNs - Physics - Teaching

Code soon-to-be released - CV

Kupcinet-Getz Scholar — Computational Biochemical Diffusion - Jun 2025 – Aug 2025 - Selected as one of 22 worldwide Kupcinet-Getz International School scholars; investigated stochastic differential equations for enzyme reaction-diffusion systems with Brownian motion, chemical kinetics, spatial heterogeneity, nonlinear biochemical feedback oscillations, molecular self-organization, and chaotic dynamics; developed PINN/neural-ODE methods, custom Euler-Maruyama, Runge-Kutta, and Gillespie SSA solvers, plus GNN/transformer/UMAP analysis of trajectories.

Biochemical Diffusion - PINNs - Neural ODEs - Stochastic Simulation

Code soon-to-be released - CV

Advanced Organic Synthesis — Buchwald Lab - Dec 2025 – May 2026 - Supported development of CuH-catalyst chemistry for selective asymmetric methylation and alkylation of vinyl boronate esters; used 1H-NMR and chiral-HPLC to connect catalyst structure and activity, expand substrate scope, and scale syntheses from 0.1 mmol to multi-gram quantities.

Organic Chemistry - Synthesis - Catalysis - NMR

Code soon-to-be released - CV

Sepal AI / Mercor — Chemistry LLM Evaluation and Scientific Data - Sept 2024 – Feb 2025 - Developed model-training, testing, and evaluation data for LLM chemistry workflows, including graduate-level retrosynthesis, reaction mechanisms, and method-selection tasks; designed grading rubrics, chemical-validity constraints, prompt-engineering workflows, error analysis, and structured fine-tuning/evaluation formats.

LLM Evaluation - Chemistry - Data - Prompt Engineering

Code soon-to-be released - CV

Computational Organic Chemistry Pipeline — DFT to kMC - Aug 2023 – Feb 2024 · Lumiere Research Inclusion Foundation - Engineered an automated multi-scale simulation pipeline linking DFT calculations to stochastic kinetic Monte Carlo models; screened 20+ transition states and 500+ reaction trajectories across 5 solvent pathways, reaching a 22% predicted yield optimization for a cytotoxic target while working 1:1 with a Brown University PhD mentor.

DFT - kMC - Computational Chemistry - Reaction Dynamics

Code soon-to-be released - CV

MIT 6.1200[J] Teaching Fellow and Grader - Jan 2026 – Present - Selected after receiving an A+ and ranking top 1% in MIT EECS’s largest foundational theoretical CS course; supporting a 250+ student cohort through recitations, office hours, proof-intensive grading, midterms, and finals in logic, graph theory, recurrences, asymptotics, and cryptography.

Teaching - Algorithms - Discrete Mathematics - Proofs

Code soon-to-be released - CV

Formula SAE — Low Voltage and Software Team - Jan 2025 – May 2025 - Worked on PCB systems and embedded software for a Formula SAE electric racing vehicle, collaborating on sensor-network integration, low-voltage microelectronics, firmware development, Altium Designer workflows, and C/C++ implementation.

Embedded Systems - PCB - C/C++ - Engineering

Columbia Formula Racing - CV

Chemistry Olympiad Team Captain and Club President - Sept 2020 – Jun 2024 - Led ACS Chemistry Club and Chemistry Olympiad Team; mentored a 50-member team on advanced olympiad problems, delegated 15 senior members across organic, inorganic, analytical, and physical chemistry, organized 15 two-hour laboratory training sessions, and helped 35 members place in top rankings across regional, national, and international competitions.

Leadership - Chemistry - Mentoring - Olympiads

Code soon-to-be released - CV

Varsity Volleyball Team Captain - Sept 2020 – Jun 2024 - Served as team captain and primary setter; coordinated weekly training sessions, initiated team-bonding activities, and organized service activities including beach clean-ups.

Leadership - Athletics - Teamwork - Community

Code soon-to-be released - CV

MIT BioMakers Lab Initiative — Dry-Lab Member - 2026 – Present - Dry-lab member in the MIT BioMakers community, contributing computational, modeling, and project-infrastructure support for biology-facing student projects and interdisciplinary biomaking work.

BioMakers - Dry Lab - MIT - Community - Biology

MIT BioMakers - CV