1 of 10

Navigating the AI Revolution in Industry

Jainam Shah �Manager Research Engineer ��

Bachelor’s in Computer Science | India

Master’s in Computer Science | New Jersey Institute of Technology | USA

Worked in 4 different startups, with the last being acquired by Thomson Reuters

Relocated to India since a year now

2 of 10

How we work�@ Labs

3 of 10

Work

Working with Large Language Models (LLMs) is a major part of our job
Team of research scientists and engineers collaborating with SMEs (mostly editors and lawyers) on content lifecycle
Focusing on enabling content journey from raw to remarkable with AI
Generating summaries, rule-based document modification, drafting, data point generation, and more
Company believes in AI as a tool to transform current processes, making them more accurate and faster, with human review at the end

4 of 10

Process

Understanding problem / current implementation

Gathering data (build Gold datasets)

Clearly defining Code + LLM Strategies

Raw implementation and SME loop / Prompt engineering

Evaluation with SME or gold data and in some cases using LLM

Optimization

Deployment

Monitoring and updating as per the performance over the time

5 of 10

Research�Exploring and implementing the latest AI trends and techniques to stay at the forefront of innovation

Tool Generation Leveraging LLMs to create tools and utilities that streamline workflows

Metaprompting Developing effective prompts using LLMs

Fine-tuning Adapting LMs to specific domains and use cases for improved performance

Model Evaluation Assessing the performance, accuracy, and reliability of LLMs in real-world scenarios based of our data

Agents�Working with various models and creating agentic workflows

6 of 10

AI Transformation Hurdles

Legacy technology and workflows can be difficult to understand and adapt to newer processes

Prompt engineering is a major aspect of AI adoption, requiring domain understanding and engineering collaboration

Hallucinations and inconsistent outputs

Handling token consumption and pricing

Testing models to find the best fit for each use case

Significant engineering work required for pre and post-processing in content modification

Keeping pace with the rapidly changing industry and new research

7 of 10

Shifts in Ways of Working

Access to LLMs for code generation has significantly impacted engineering workflows
Learned not to force LLMs when rule-based approaches suffice
More time spent understanding use cases and solution design rather than development
Strong engineering knowledge remains crucial, but solution development is faster-paced
Fierce competition from startups with novel ideas and solutions
Roles have been revamped to align with current market needs
Increased presence of scientists and engineers to adapt to these changes

8 of 10

Key Learnings

Embrace AI as a tool to enhance and streamline processes, not as a replacement for human expertise
Foster collaboration between domain experts, research scientists, and engineers
Stay updated with the latest industry trends and research to remain competitive
Adapt roles and skills to align with the evolving market demands
Focus on understanding problem statements and designing effective solutions
Maintain a strong foundation in engineering principles while leveraging AI capabilities
Be prepared for rapid changes and continuously learn and grow

9 of 10

In the not too distant future

AI will continue to transform various industries, with a focus on enhancing efficiency, accuracy, and decision-making
Increased adoption of AI-driven solutions in all industries.
Growing demand for professionals with AI expertise and domain knowledge to drive industry-specific applications
Emergence of new roles at the intersection of AI, engineering, and specific industries
Emphasis on developing explainable and interpretable AI models to ensure transparency and trust
Continuous evolution of AI technologies, with a focus on making them more accessible and user-friendly for businesses of all sizes