As a Software Engineering Intern at Outlier AI, I contributed to the development of a specialized LLM for programming by designing and implementing key algorithms. I developed and optimized parsing and tokenization methods, which improved the model's processing speed by 20%. I also enhanced the model's accuracy by 10% through the refinement of its training data pipeline and integration of advanced validation techniques. Additionally, I established rigorous testing protocols and code review practices in Python and JavaScript, achieving a 95% accuracy rate and accelerating task completion by 30%.

I teach English and Violin to over 120 children in Coque, one of the most violent neighborhoods in Brazil. I have spent time over my summer holidays teaching classes since 2018 and regularly teach online lessons throughout the year. It has given me an appreciation for philanthropic work, and I hope to run my own organisation like OCC one day.

Using Simulation-Based Inference and Machine Learning to model strong gravitational lensing in galaxy clusters. My work aims to make galaxy cluster lensing more accurate and much faster — currently it takes days to model one galaxy cluster.

Conducted molecular biology experiments to uncover Trypanosoma pathogenesis in human cells. Developed bioinformatics algorithms in R for molecular docking, demonstrating Metformin's therapeutic potential on neural tissues, publishing two research papers and contributing to three new research proposals. Paper published in the Biomolecules journal (Bharati, J., Kumar, M., Kumar, N., Malhotra, A., & Singhal, P. C. (2023). MicroRNA193a: An emerging mediator of glomerular diseases. Biomolecules, 13(12), 1743. https://doi.org/10.3390/biom13121743).

Investigated the olfactory cortex of fruit flies, translating biological mechanisms into a novel artificial neural network model in Python, reducing catastrophic forgetting by 20%. Achieved a 15% improvement in memory retention over existing algorithms through extensive empirical testing. Paper is currently under peer review for publication in the Science journal.

I am interested in cognitive psychology, decision making, heuristics, belief systems and truth. Heuristics or mental shortcuts (of which there are about 50, including endowment, attribution, hindsight, recency, self serving, availability, anchoring, representativeness, familiarity, bandwagon, etc) result in blind spots and biases. Understanding how the brain architecture, neuronal connections and their interplay results in thoughts may have an application in the mitigation of biases in computer decision making.