## **Team Lead, AI R&D at AutomationEdge, Pune**

> January 2019 - December 2021

As a Team Lead in AI Research and Development at AutomationEdge, I played a pivotal role in shaping the development of the world's first Hyper-automation platform with Autonomous AI for Business & IT. My contributions and achievements in this role include:

**Natural Language Understanding Module:** I spearheaded the development of a cutting-edge Natural Language Understanding (NLP) module that demonstrated an impressive accuracy rate of 92%. This module was designed to comprehend IT domain-specific language, enhancing our platform's capabilities to understand and respond to intricate technical queries.

**Enterprise Search Engine:** I led the design and implementation of an enterprise search engine that showcased intelligent data indexing capabilities. This innovative solution efficiently gathered and organized data from various sources, facilitating faster and more accurate information retrieval.

**Team Lead:** Managing a talented team of seven professionals, I oversaw the creation of an exceptional NLP system. This system proved highly efficient, automating approximately 60% of IT ticketing processes for organizations facing ticket volumes of 100k or more annually. This achievement significantly reduced operational overhead and improved overall efficiency.

My tenure as Team Lead at AutomationEdge was marked by a commitment to innovation, a focus on harnessing the power of AI for practical solutions, and effective team leadership. These experiences have honed my expertise in AI, NLP, and automation, contributing to my strong foundation in the field of technology.

These accomplishments, along with my passion for AI and automation, reflect my dedication to driving technological advancements and solving complex business challenges. I invite you to explore my personal website and resume to learn more about my contributions and expertise in the realm of AI and Cyber Security.

AutomationEdge

## **Founder at ICyberSol, Bengaluru**

> April 2017 - December 2019

As the Founder of ICyberSol, I embarked on a visionary journey to establish a cutting-edge cybersecurity firm. My entrepreneurial endeavors and achievements in this role include:

**Strategic Leadership:** I provided visionary leadership by establishing the organization, defining its strategic direction, and executing operational plans to achieve our ambitious goals. My ability to align the team with our overarching mission was instrumental in our success.

**Intelligent Intrusion Detection System:** I conceptualized and personally led the development of one of the industry's pioneering Intelligent Intrusion Detection Systems (IDS). This ground-up project yielded an impressive accuracy rate of 93%, marking a significant advancement in cybersecurity technology. Our IDS played a crucial role in safeguarding against cyber threats.

**Market Analysis and Customer Acquisition:** I conducted thorough market analysis, identifying critical opportunities and challenges. Leveraging this knowledge, I meticulously developed our cybersecurity product, aligning it with market demands. My efforts extended to crafting a sustainable business model that ensured our long-term viability. Within a span of just two years, I successfully onboarded eight B2B customers, establishing ICyberSol as a trusted partner in the cybersecurity landscape.

My role as Founder at ICyberSol encapsulated my unwavering commitment to innovation, strategic thinking, and the pursuit of excellence in the cybersecurity domain. These experiences have fortified my expertise in cybersecurity strategy, product development, and market analysis.

I invite you to explore my personal website and resume to delve deeper into my journey as an entrepreneur and my dedication to creating pioneering solutions in the field of cybersecurity.

ICyberSol

In the dynamic realm of cybersecurity, my work in 2018 took on a highly technical and ambitious dimension, culminating in the creation of a Hybrid Intrusion Detection System. Over a meticulous 1.5-year development period, I engineered a solution that expertly combined machine learning techniques to bolster intrusion detection capabilities. Notably, this system extended its purview to breach detection as well.

**Approach:**

The development journey commenced with a strategic approach rooted in the fusion of time series analysis and deep learning techniques. Leveraging the versatility of time series algorithms, we established a robust foundation for capturing temporal patterns within network data, crucial for recognizing suspicious behaviors.

This foundation was further fortified by the incorporation of deep learning methodologies. Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) were implemented to extract intricate features from network traffic data, enabling a deep understanding of network dynamics and patterns.

**Achievements:**

Our efforts yielded remarkable results. The system consistently achieved an accuracy rate of an impressive 93%, surpassing industry benchmarks and showcasing the prowess of our hybrid approach. This high degree of precision ensured that even the subtlest anomalies were promptly identified.

One of the most notable achievements was the system's rapid inference speed. Through careful optimization, we reduced the time taken for data processing and determination to an astonishing 1/8th of a second. This remarkable speed was pivotal in our endeavor to safeguard network integrity in real-time, swiftly responding to potential threats.

**Inference Acceleration:**

To further enhance inference performance, we harnessed the power of the Movidius Neural Compute Stick (NCS). By offloading inference tasks to this specialized hardware, we not only accelerated processing but also reduced the computational load on the system, ensuring efficient and rapid threat identification.

**Data Collection:**

To underpin our system's efficacy, I embarked on a comprehensive data collection endeavor, dedicating six months to construct a private network environment. This meticulously curated network served as the foundation for our data collection efforts, enabling the generation of authentic, real-world data for training and validation purposes. The inclusion of this authentic data contributed significantly to our system's unparalleled accuracy and performance.

In conclusion, this technical accomplishment underscores my unwavering commitment to technological innovation, precise execution, and rigorous data-driven methodologies. Our hybrid intrusion detection system, empowered by time series algorithms, deep learning techniques, and hardware acceleration, represents a remarkable advancement in the field of cybersecurity. It is a testament to our dedication to enhancing security practices and safeguarding digital ecosystems through cutting-edge technology and comprehensive data analysis.

Intelligent Intrusion Detection System

Looking for Co-op and Full time opportunities

Team Lead, AI R&D

INDIA PRIME TOP 100 Professors  Researchers & Teachers

India Prime Icon Award

Top 100 Innovators and Influencers

Young Achievers Award

This specialisation is a completion of following 5 courses:

1. Convolutional Neural Networks
2. Neural Networks and Deep Learning
3. Sequence Models
4. Structuring Machine Learning Projects
5. Improving Deep Neural Networks: Hyperparameter Tuning, Regularization and Optimization

Deep Learning Specialisation

I am currently working on development of a privacy-friendly federated learning framework capable of handling Large Language Models such as Llama2 and Falcon. This framework integrates Secured MPC algorithms to guarantee high levels of security, confidentiality, and privacy throughout the federated learning process. Additionally, I have optimized the MPC algorithms, leading to a remarkable 93% reduction in execution time and a 45% decrease in memory usage per round of Federated Learning. This combination of capabilities positions the framework as a robust and efficient solution for privacy-preserving federated learning tasks involving complex language models.

Secured Federated Learning

In 2018, I spearheaded the development of a Hybrid Intrusion Detection System, a technically advanced solution combining time series algorithms and deep learning techniques. Achieving a remarkable 93% accuracy rate, the system rapidly processes network data in 1/8th of a second, thanks to optimization and hardware acceleration using the Movidius NCS. A comprehensive data collection effort, including the creation of a private network environment, contributed to the system's unparalleled performance. This accomplishment underscores my dedication to cybersecurity innovation and data-driven precision.

Software Innovator

Mentor

Lead & Co-founder

Student Ambassador

Managed the research portfolio and worked on various advanced research analytics in the Healthcare domain.

Research Manager

• Lead NLP team in developing the world's first Hyper-automation platform with Autonomous AI for Business & IT
• Developed a Natural language understanding module that understands IT domain-specific language with an accuracy of 92%
• Built an enterprise search engine that intelligently indexes different data from various sources.
• Managed team of 7 in building the exceptional NLP system that can automate nearly 60% of IT tickets for an organization.

• Led organization, established business direction, and actualized operational plans to meet goals.
• Designed and developed one of the first intelligent intrusion detection systems from the scratch with an accuracy of 93%.
• Performed market analysis, developed the product, and designed. a business model, and finally onboarded 8 B2B customers in the span of 2 years.

Founder

• Guided and worked closely with PhD students from multiple universities in the United Kingdom in developing a novel and
efficient Insider threat detection system.
• Used a combination of ML models to develop an insider threat detection model on a large-scale dataset that had a test accuracy of 97%.

Sri Harsha Gajavalli

2023

2021