Navigation
terminal
harshini@dev:~$whoami
Harshini - a bioinformatician and ML researcher, now doing an MSc in Applied Bioinformatics at King's College London, after a B.Tech in Biotechnology with a Minor in Computer Science at PES University. I work where single-cell & bulk transcriptomics, ML for RNA therapeutic design, explainable AI and drug discovery meet.
Most recently a Bioinformatics Scientist at NCBS, where I built an RNA-therapeutics platform covering 41 mechanisms (submitted to ICLR 2027 and sent for patent), after a year at IISc.
harshini@dev:~$
Try terminal commands like "help", "ls", "previous", "cd consumedlog", "pwd", "contact", "clear"
What I've Been Up To
A glimpse into the milestones, collaborations, and projects that have shaped my path in research — the work, the people, and what I learned.
2026
- Started my MSc in Applied Bioinformatics at King's College London⧉↗, carrying everything below with me to London. I completed my B.Tech at PES University with Distinction in both my major in Biotechnology and minor in Computer Science, equivalent to UK First-Class Honours, alongside an honours thesis integrating in vitro and in silico approaches under Dr. Reshma S V, ranked in the top 10% of the graduating cohort among 2,800+ graduating students.
- As a Bioinformatics Scientist at NCBS⧉↗ (KoshKey⧉↗, with Dr. Tanay Bhatt⧉↗), I built a single-pass mechanism-arbitration platform spanning 41 RNA therapeutic mechanisms across 9 therapeutic goals, with 19 mechanism-specific features that rank candidate oligos. I benchmarked 7 ML models across 12 controlled experiments on 165,237 RNase H, splice-switching and siRNA rows, backed it with 95 tests, and the work is now submitted to ICLR 2027⧉↗ and sent for patent.
- At NCBS, investigated scleroderma-associated fibroblast heterogeneity⧉↗ using 10x Genomics single-cell RNA-seq across wild-type and disease-phenotype mouse populations at multiple age points. Performed batch-aware Seurat integration, fibroblast subpopulation characterisation, pathway and gene-signature analysis, and Monocle3 pseudotime trajectory inference; analysed cell–cell communication using CellPhoneDB and contributed to primer design for follow-up experiments.
- Wrapped up a year at IISc⧉↗ with Dr. Payel Roy, analysing human and mouse T cells in atherosclerosis and stimulation-resolved immune cohorts. I received the Research Excellence Honour for my contributions among the final-year cohort at PES University, and our manuscript is submitted to the Journal of Immunology, the official publication of the American Association of Immunologists (AAI).
- Built GenomeSearch⧉↗, an end-to-end genomic sequence search engine integrating k-mer indexing, hash-based retrieval, trie prefix search, memory-optimised Smith–Waterman alignment, DAG-based workflow scheduling, concurrent multi-process/thread execution and a normalised PostgreSQL database, benchmarked for runtime, memory and scalability on real NCBI RefSeq genomes.
- Conducted a comparative genomics study of Apis mellifera⧉↗ to investigate the genetic basis of social behaviour, including foraging, division of labour, caste differentiation and pheromone-mediated communication. Analysed genome annotations and homologous sequences using NCBI Genome Browser, BLAST and InterProScan; performed multiple-sequence alignment with MAFFT and phylogenetic analysis using MEGA and RAxML to assess evolutionary conservation and divergence of behaviour-associated genes.
- Performed a structure-based virtual screening study of MMP-7⧉↗ covering 50 potent inhibitors from ChEMBL (CHEMBL4073) against PDB 1MMQ using AutoDock Vina, evaluating binding affinity, catalytic Zn²⁺ chelation geometry and ligand efficiency. Characterised the metal-binding active site, developed dual-criterion hit-ranking using Pareto-front analysis, and identified CHEMBL41497 (−9.25 kcal/mol) as a top candidate with bidentate-like Zn²⁺ coordination. Built reproducible docking workflows and interactive tools using a web platform, Streamlit, Docker and Google Colab.
- Developed comparative ML pipelines for disease prediction⧉↗ across cardiovascular, metabolic, oncological and neurodevelopmental diseases using ANN, KNN, Logistic Regression, SVM, Random Forest, Gradient Boosting and XGBoost. Engineered features, handled missing values and outliers, optimised hyperparameters using GridSearchCV/RandomizedSearchCV and stratified cross-validation, and evaluated generalisation using precision, recall, F1-score and ROC–AUC. Improved diabetes-prediction test accuracy from 70.1% to 75.2% and built reproducible Python workflows using Pandas, NumPy, Scikit-learn, XGBoost and visualisation libraries.
- Engineered a JSON-driven laboratory protocol validator⧉↗ with an O(S) simulation engine to detect pipette-range violations, reagent depletion, well overflow, invalid well references and missing controls. Benchmarked fault detection using seeded synthetic protocols and 95% Wilson confidence intervals; validated 4 Opentrons workflows with 0 false positives. Built a 0-dependency JavaScript web app featuring a 96-well plate simulator, protocol composer and real-time reagent-state visualisation.
- Built BodyTrack Atlas⧉↗, an interactive index of 500+ open imaging datasets across 11 model organisms, integrating curated OpenMedLab and TCIA resources with verified source links, citations and suggested analysis pipelines. Developed anatomy-linked dataset discovery with interactive body maps, 3D CT/microCT models and a client-side image/CSV analysis workbench for privacy-preserving data exploration.
2025
- Designed an interpretable machine-learning framework for cardiovascular risk prediction from longitudinal cohort data⧉↗, integrating clinical, lifestyle, laboratory and immunological features to model heterogeneous disease-risk profiles. Combined Logistic Regression, SVM, Random Forest, gradient boosting, neural networks, Random Survival Forests and DeepSurv, incorporating single-cell transcriptomics, TCR repertoires, APOB-specific immune responses and clinical biomarkers through dimensionality reduction and representation learning. Integrated clinical, molecular and immune features into unified hazard and risk estimates; evaluated performance, calibration, clinical utility and generalisability using C-index, integrated Brier score, time-dependent AUC, calibration curves, decision-curve analysis, cross-cohort transfer and feature-set ablation. Aligned longitudinal clinical and immune trajectories to account for irregular sampling, evolving baselines and dynamic cardiovascular risk.
- My paper⧉↗, "A Multi-Method Interpretability Framework for Probing Cognitive Processing in Deep Neural Networks across Vision and Biomedical Domains", was selected for two NeurIPS 2025 Workshops — SPIGM⧉↗ and Reliable ML from Unreliable Data⧉↗, held in California, USA. NeurIPS is one of the top machine learning conferences in the world, and seeing a project I’ve been working on since 2023 reach this stage has made every bit of effort worth it.
- My paper⧉↗, “PathSymetic: Neuro-Symbolic Causal Hypothesis Generation for Mechanistic Pathway Discovery in Genomic Systems”, was selected for the AAAI 2026 SPARTA Workshop⧉↗, Bridge LMReasoning⧉↗, and AI4Research Workshop⧉↗, taking place in Singapore on January 2026. This work explores neuro-symbolic causal hypothesis generation for mechanistic pathway discovery, bridging machine learning and structured biological knowledge.
- Got a Summer Research Internship at the Indian Institute of Science (IISc) to work under Dr. Payel Roy⧉↗. If somebody told the 17-year-old me that I’d get the opportunity to do research here, I wouldn’t have believed them. Dreams do come true. Working with Dr. Roy has been one of the most formative experiences of my journey, teaching me the importance of breaking complex problems into smaller, meaningful pieces and to look at research with patience, depth, and genuine curiosity.
- Presented my first paper paper⧉↗, “Evolving Innovative Process for Bioethanol Production from Tropical Fruit Residues Integrated with Biomaterial Catalyst and Saccharomyces cerevisiae Bioagent,” at the International Conference on Recent Advances in Environmental Science (ICRAES)⧉↗, MSRIT, Bangalore, under the guidance of Dr. Dinesh M. S.⧉↗and Dr. Kavitha S. H.⧉↗. The work was published in the prestigious Taylor & Francis journal⧉↗.
- Presented my second paper⧉↗, “Deep Learning-Based Classification of Plants and Detection of Red Palm Mite Infestation from Leaf Images,” at the Fourteenth International Conference on Contemporary Engineering and Technology (ICCET)⧉↗, Chennai, Tamil Nadu, under the guidance of Professor Sunil K.C⧉↗.
- Academically, I completed my Minors in Computer Science this year and I managed to be in the top 5% of the batch across all the subjects including Operating Systems (OS), Data Structures and Algorithms (DSA), Database Management Systems (DBMS), and Design and Analysis of Algorithms (DAA), something I worked hard for while balancing my research work.
- Around the same time, another work⧉↗, “Machine Learning-Augmented Simulation of Bioethanol Production from Vegetable Waste, Lignocellulosic Biomass, and Marine Algae Using MATLAB,” was accepted at the ACMS 2025 Conference⧉↗. Over time, I’ve developed a strong foundation in Machine Learning (ML) and Computational Modeling, and consistent practice has helped me apply its methods to niche interdisciplinary areas such as biofuel process modeling and simulation.
- I was selected for the Centre for Innovation and Entrepreneurship (CIE)⧉↗ program from a pool of over 1,400 applicants. Our group, comprising myself, Sameeksha, Mallikarjun, Nischal, Anish, Achal, and Rishab, received a special mention for Best Insights and Customer Discovery. As part of the program, we developed innovative product ideas and presented them through a Business Model Canvas (BMC)⧉↗ across multiple evaluation levels.
- I’ve also been collaborating with Nikhita (Ph.D. scholar, IISc) on T-cell research, alongside Dr. Dipankar Nandi⧉↗ and Dr. Payel Roy, where I contribute to the computational analysis of immune response data and modeling of T-cell signaling networks.
- I built my Capstone team with Subhangi Swain and Tejal Mahesh, under the guidance of Dr. Reshma S V.⧉↗, and together we began ideating for our research proposal⧉↗, “Phytochemical-Mediated Quorum Quenching Activity of Premna seratifolia against Antibiotic-Resistant Pseudomonas aeruginosa”. We were certain about focusing on Antimicrobial Resistance (AMR) as our main area of interest, drawing from our prior experience in plant tissue culture. The topic appealed to us because it beautifully bridges both wet-lab experimentation and dry-lab computational analysis. We were really determined to work under Mrs. Reshma, and it's not easy getting her as the guide for Capstone because almost the entire batch would want to work under her. We had proved ourselves to be worth taking the risk and having us under her.
- I also worked with my capstone team mates on proposing a book chapter⧉↗ titled “Harnessing Bioinformatics and AI for Microbiome-Targeted Discovery of Plant-Derived Therapeutics” under the guidance of Dr. Reshma S V. The chapter has been accepted for publication by IGI Global⧉↗, and the experience offered valuable insight into bioinformatics-driven research and interdisciplinary scientific writing.
- I had the opportunity to attend a seminar hosted by Beckman Coulter⧉↗ Life Sciences in Bangalore, where they introduced a new set of Flow Cytometers for the first time. The event was attended by leading industry experts and prominent academicians, providing valuable insights into the latest advancements in flow cytometry technology.
- Got really interested in Big Data Analysis and started working on the paper “TriggerNet: A Novel Explainable AI Framework for Red Palm Mite Detection and Multi-Model Comparison and Heuristic-Guided Annotation” which is currently available on arXiv⧉↗. This got me interested in the field of Machine Learning too, where applying ML and deep learning algorithms on biological systems and later analysing the results, trying my hands on anything and everything, and seeing that interests me the most. I plan to develop it into a comprehensive explainable AI framework for which the implementations will be open-sourced soon.
- Wrote the review⧉↗“Precision Agriculture-Based Pest Detection Solutions Using AI Algorithms Simulated with Complex Logic Gates Integration and 3D Bioprinting,” presenting a forward-looking framework that benchmarks AI models (CNN, SIFT, Random Forest) for rapid pest identification, while exploring logic-driven decision systems and 3D-bioprinting-inspired pest-resistant plant tissue design to enable resilient and sustainable crop protection.
- Recognized at the Biotech Revolutions Hackathon⧉↗, conducted at PES University, Bangalore, for developing an innovative biodegradation-based solution to reduce pesticide residues in fruits and vegetables. Collaborated with Preetham, PhD scholar at PES University, to design a model⧉↗ integrating enzymes (laccase, peroxidase, carboxylesterase) and probiotics (Lactobacillus, Pseudomonas) for sustainable pesticide breakdown and safer agricultural practices for the same.
- This year, I had the opportunity to serve as a reviewer for the AAAI Conference, 2026⧉↗ on Artificial Intelligence (AAAI 2025) and the NeurIPS 2025 Workshops⧉↗, where I evaluated submissions focused on machine learning, interpretability, and computational biology. It was a deeply insightful experience that strengthened my understanding of research evaluation and the peer-review process in top-tier conferences.
- Replicated the Compass⧉↗ framework, one of the most detailed and phased computational pipelines for both single-cell and bulk-cell data analysis. The project⧉↗ involved reproducing each stage from preprocessing and modeling to interpretability and reporting to ensure the workflow’s reproducibility and transparency. Used single-cell analysis with Compass to analyze metabolic flux changes, reconstruct cellular differentiation trajectories, and study immune cell behavior. The complete implementation and analysis are available on my Github⧉↗ page.
- Implemented a detailed RNA-seq analysis pipeline⧉↗ from raw FASTQ files to TPM quantification and differential expression analysis. The workflow utilized FastQC, Fastp, MultiQC, and Trimmomatic for quality control and trimming; STAR, Hisat2 and Subread for alignment and quantification in quant mode; and SAMtools for BAM file processing. Post quantification, performed normalisation and differential expression analysis using Deseq2 and EdgeR supported by R libraries such as ggplot2, EnhancedVolcano, and clusterProfiler for visualisation and functional enrichment analysis. Additional utilities like LLAMA and BEDTools were used for genomic feature assessment. The entire pipeline is documented and available on my GitHub for reproducibility.
- Authored a review paper⧉↗ as first author titled “Valorization of Diverse Tropical Fruit Residues for Bioethanol Production: A Review of S. cerevisiae Strains and Simulation Models”, presenting a computationally-driven synthesis of 150+ studies on enzymatic/acidic hydrolysis, fermentation kinetics, yeast strain selection, and ripeness-linked sugar variability, supported by MATLAB simulations for multi-feedstock tropical fruit bioethanol potential, completed under the guidance of Dr. S. H. Kavitha.
- Started as a question to myself: can renewable fuel actually come from the waste we overlook every day? That thought sparked my next first-authored review paper⧉↗, “Developing a Sustainable Process for Bioethanol Production from Vegetable Residues, Lignocellulosic Biomass, and Marine Algae Integrated with Saccharomyces cerevisiae Bioagent”. The work expanded organically; what began with fruit waste soon grew into exploring vegetable residues and biomass feedstocks, focusing on fermentation kinetics, substrate behavior, and sustainable conversion.
- Continuing from the first two ideas, I first authored another review paper ⧉↗, “Expanding the Bioethanol Potential: Quantification, Qualification, and Applications of Diverse Temperate Fruit Waste" analyzing bioethanol feasibility from grape pomace, watermelon rind, sapota peel, passion fruit peel, grapefruit peel, pear, Mousambi, longan, blueberries, apples, pomegranates, and other residues.
- Performed eight-layer structural and functional annotation of Neuropilin-2 (NRP2; UniProt O60462)⧉↗ using 20+ bioinformatics tools, integrating BLAST, MSA, HMM-based topology prediction, domain annotation, proteolytic cleavage mapping, PDB structure analysis and STRING/IntAct interactomics to investigate semaphorin–plexin and VEGF signalling. Developed an interactive web platform and Colab notebook featuring sequence analytics, hydropathy plots, UPGMA phylogenetics, interactome visualisation and motif/PTM scanning.
- Engineered an end-to-end analytics pipeline for Patient-Reported Drug Review Analytics & Ranking⧉↗ across 215,063 patient-reported drug reviews, 836 conditions and 3,654 drugs, cleaning 212,698 records and correcting 1,171 label contaminations. Developed shrinkage-based rankings across 8,586 drug–condition pairs, integrating rating-distribution analysis, temporal drift assessment and precomputed sufficient statistics for real-time sensitivity analysis in a zero-backend application.
- Built DataSphere⧉↗, a polyglot database system integrating MySQL, MongoDB and Neo4j to manage e-commerce transactions, product data and customer–product relationships. Implemented 3NF/BCNF schema normalisation, SQL analytics, query optimisation and graph-based recommendations; benchmarked cross-database performance using median, p95 and p99 latency, indexing efficiency and transaction behaviour.
- Analysed the 2.10 Å crystal structure of anti-AT1R nanobody AT118i4h32 (PDB 7T83)⧉↗, annotating CDR/framework regions, disulfide bonds and crystallographic quality. Ran a 50 ns GROMACS molecular dynamics simulation using ff14SB/TIP3P at 300 K and 0.15 M NaCl, evaluating RMSD, radius of gyration and residue-level RMSF; validated simulated flexibility against experimental B-factors and developed interactive structural-analysis tools using PyMOL, Streamlit and Google Colab.