About me

I’m a MLE at AWS Annapurna Labs architecting PyTorch & JAX frameworks to "Just Work" on AWS Trn/Inf, cross-functionally driving PJRT, compiler & runtime development. Spanning the neuron stack, I optimize distributed model performance via graph fusion, TP, sharding & kernel design, while triaging compiler-runtime gaps, enabling cloud AI workloads. My expertise stems from deploying LLM training, fine-tuning & inference on custom accelerators at SambaNova and a funded MS at CMU RI SCS with CV, DL & ML depth and SWE Intern experience at JBT corp. At CMU, I cultivated expertise in creating practical AI solutions that align with business principles while effectively serving client needs. My experience spans 3 diverse labs, coordinated with Pitt and UPMC physicians, sponsored by DoD, NSF and DARPA, emphasizing 3 industry-scale AI challenges: 1. Transparency, deployability and resource limitations for AI adoption: Heuristics-guided explainable AI; 2. Precision inadequacy in automated robotic interventions: Physics-informed gen AI; 3. Vast unlabelled and unstructured training data: Contrastive unsupervised representation learning. Summer of 2023, interning at JBT, collaborating with ifm, Oppent, and OSU, I integrated an O3R camera with autonomous vehicle navigation-vision C++ stack on edge VPU, enabling obstacle detection, generating requirements, reducing stack lag and streamlining verification & validation protocols while leveraging agile practices, version control and CI/CD pipelines. Before CMU, several global collaborations with universities: UWaterloo, NUS, ICL and CUHK (exchange study) exposed me to designing industrial automation software: A clinical diagnosis and experimental analysis app to aid ophthalmologists and CV-DL frameworks to tackle automation delays in medical imaging (Graph networks) & latency in computer-assisted surgery (Paced curriculum learning + knowledge distillation). My introduction to CV, DL & ML was at IIT-M, building sketch query 3D CAD model CNN retrieval systems and honed my skillset at Origin Health Pte. Ltd., prototyping in-house Python GUI and fetal screening CV-algorithms that piqued VCs. Moreover, I garnered this passion initially, devising a DL-based EMG-controlled CPM machine, as an undergraduate engineer at my university robotics club. Given my strong academic background, accolades, 5+ years practical experience developing industry solutions in diverse CV, DL & ML global ventures, with 10+ publications, combined with my demonstrated track record of soft skills, I am an excellent candidate for a relevant position.