Abhishek Agarwal Email & Phone Number

Santa Clara, California, United States

I am working on the networking fabric to enable scale-out capabilities for Tenstorrent hardware.

San Diego, California, United States

• Worked on writing kernel-mode drivers to enable communication between different sub-systems on an SoC. Was also involved in system-wide initiatives to better align the software with organizational goals.
• Developed and contributed to IPC drivers for Windows, Qualcomm RTOS, and Zephyr RTOS, for 6 mobile, compute, and automotive chipsets.
• Designed and prototyped a Distributed Computing Framework for running services in a distributed fashion on an SoC such as a distributed NoSQL database and a Capability Advertisement service.
• Collaborated with cross-functional teams and addressed architectural differences to enable IPC for RISC-V based subsystems.
• Led the design and implementation of a highly scalable system control framework compliant with ARM’s SCMI specification.
• Implemented a memory allocation scheme to reduce memory corruption; designed experiments to make data-driven decisions.

San Diego, California, United States

• Worked with the Core Platform Kernel team on Qualcomm Real-Time Operating System
• Analyzed memory and performance trade-offs of deploying new process loading and linking schemes on the hexagon DSP
• Implemented a prototype of a run-time loader (rtld) library utilizing GNU hashing, resulting in 50% faster run-time linking
• Presented a collection of best practices on creating shared objects for efficient loading and linking of dynamically linked processes
• Developed a profiling tool to auto-generate binaries based on user inputs and collect build and run-time performance metrics
• Integrated the tool with the existing build system and regression testing framework to conveniently stress test new features

Bengaluru, Karnataka, India

• I was responsible for building Fraud Risk management framework for the credit offerings (Buy Now Pay Later and EMIs) of the FinTech division. I also led the development of Credit risk framework (Underwriting and.
• Collaborated to design and launch the first version of transaction authorization framework for real-time authorization and monitoring of ∼2M monthly Credit transactions
• Localized 5X risky customers through network graphs and established routine pruning processes
• Built databases hosting historical credit performance and shopping trends, accelerating routine analyses by >90%
• Developed scripts to scrape data from the Internet such as device information, IP databases, etc., to augment internal databases for enhanced fraud managementFlipkart Wholesale Easy Credit (Working Capital credit for.
• Led the development of credit underwriting and fraud mitigation strategies for the new wholesale business launch

Mumbai Area, India

• I was responsible for building data-driven strategies for insurance products (includes life and general insurance). I worked on launching campaigns and supporting business strategies through statistical.
• Developed a logistic regression model to identify customers with the lowest medical risk by predicting the probability of a claim
• Identified customers eligible for the offer on the basis of propensity, persistency, medical, and financial assessments
• The program minimized the documentation required for underwriting and reduced the policy issuance time by 90%Profiling of Life Insurance customers:
• Designed a clustering-based framework to detect outliers in insurance sales using t-SNE and Partitioning Around Medoids
• Developed an R-shiny app to dynamically cluster and visualize customers’ attributes as per user requirements using K-meansFinancial planning of 0.4 billion USD Life and General Insurance business:

Pune Area, India

• Reinforcement Learning (RL) of Optimal ControlI assessed the viability of using RL to replace conventional control systems with adaptive model-free closed-loop control on the following two use cases. The research work.
• Trained an ’Artificial Pancreas’ capable of maintaining normal glucose levels with minimal interference in patient’s lifestyle
• Simulated Q-learning, Dyna-Q, Q(λ) and Policy Gradient on two meal models and experimented with multiple reward functions
• Formulated evaluation metrics and tuned parameters in presence of realistic disturbances in the amount and timings of mealOptimization of Fuel Consumption in a Hybrid Electric Vehicle (HEV)
• Trained a controller to optimize fuel consumption by learning optimal load distribution between the engine and electric motor
• Implemented Q-learning and Q(λ) on an HEV power-train model and used ARMAX time-series model for partial state estimation

Master Of Science - Ms, Computer Science

GPA: 3.9/4.0 Specialization: High-Performance Computing Coursework: Advanced Programming Parallel Programming Algorithms Databases Cloud.

Bachelor Of Technology - Btech, Chemical Engineering, 8.5/10

Activities and Societies: Dramatics Club, Chemineers SocietyCoursework: Computer Science: Natural Language Processing (NLP), Fundamentals.