Nov 2023 – Feb 2025 · 1 yr 3 mos · Bengaluru, Karnataka, India · On-site

• Spearheaded fine-tuning , alignment and production-readiness of India's first indigenous LLM krutrim-spectre-v1 and v2, and krutrim chat assistant.
• Led the development of Enterprise applications at Krutrim, including Agent builder platform and contact Center AI, driving efficiency and innovation.

Feb 2022 – Feb 2024 · 2 yrs

• AIM Leaders Council is an invitation-only forum of senior executives in the Data Science and Analytics industry. Members are selected based on their contribution to the analytics ecosystem in India and their achievements in the industry. Members are top management personnel, entrepreneurs, and technical heads from leading enterprises in India that work extensively on Data Science and AI.

Dec 2021 – Jan 2023 · 1 yr 1 mo

• Forbes Technology Council Is An Invitation-Only Organization For Senior-Level Technology Executives.Members are tech leaders and executives — CEOs, CIOs, CTOs, and others — selected for the council based on their deep knowledge and diverse experience in the industry.

Feb 2021 – Oct 2023 · 2 yrs 8 mos · Bengaluru, Karnataka, India

• Salesken leverages Conversational AI to revolutionise sales effectiveness, empowering sales teams to achieve exceptional results.
• Extensively used latest advancements in LLM Research (Alignment, fine-tuning, Quantization, Distillation), incorporating state-of-the-art techniques into Salesken.ai's data science workflows. Platform utilises domain-aligned finetuned & Quantized versions of Llama , MPT ,Red Pajama, Vicuna etc, while still leveraging the previous generation Language models like GPT-2, T5, BART etc as ancillary models to the LLM pipeline.
• Key highlights of the platform:
• ~ Real-time call insights: Identifying key moments in sales calls for better outcomes.
• ~ Post-call analysis: Reviewing recordings, transcripts, and analytics to improve performance.
• ~ Lead scoring: Prioritizing engagement with sales calls to focus on high-potential leads.
• ~ Pipeline health analysis: Optimizing conversion rates and decision-making.
• ~ Sales playbook: Providing winning sales techniques and step-by-step guidance.
• ~ Highly accurate transcripts: Ensuring precise performance feedback for continuous improvement.
• ~ Comprehensive and user-friendly analytics: Empowering reps to identify areas for growth.
• ~ Multi-factor lead scoring: Maximizing time and effort on the most promising leads.
• ~ Call Advisory: Receive instant feedback and guidance, enabling reps to adapt and improve their approach on the fly.
• ~ Pitch Intelligence: Helps sales teams identify effective techniques, optimize their messaging, and tailor their presentations to better resonate with prospects.

Apr 2019 – Feb 2021 · 1 yr 10 mos · Bengaluru, Karnataka

• Sr. Director - AI Research at Reliance jio Haptik. (formerly known as Reliance Jio digital services),
• Designed & architected the following features for conversational AI :
• ~Novel method for text clustering using reinforcement learning
• ~Transliteration system for indic languages
• ~ identifying Indic languages by script and in the roman form
• ~Advanced Textual entailment system , which forms the basis of an intent recognition system which scored 98f1 on SNIPS
• ~GPT-2 based Paraphrase generation system with three categories of variations : restore input sentence structure , generate surface level and semantic variations
• ~Developed the concept of intent discovery in unlabelled user queries in conversational AI .
• ~Neural question generator for generating questions from sentences & passages
• ~Multi-lingual Textual entailment
• ~Phonetic and context based sentence correction for ASR systems
• ~Question answering over Unstructured documents
• ~AI driven quality evaluation system for NLU : text style transfer, GAN
• ~Customer care agent augmentation : Observer agent for learning how the agents interact with the user
• ~Query compression for dropping noise from user queries
• ~Sentence boundary detection, punctuation restoration
• ~Reinforcement learning self-improving conversational Agent
• ~Multi agent architecture for conversations

Sep 2017 – Mar 2019 · 1 yr 6 mos · Bengaluru, Karnataka, India

• ► Spearheaded the AI R&D initiatives @ Active.ai.
• ► Designed from scratch the triniti AI stack , Cognitive QA, machine comprehension, NLU/NLG/NLI , predictive conversations
• ► contributed to prototyping of several functionalities .
• ► Solely contributed to several inventions across the stint at active.ai, among which multiple are being filed for provisional patent.
• ► My Contribution to the Machine comprehension model for Active.ai propelled the team to 9th spot on Squad 1.1
• ► Responsible for AI product roadmap
• ► Track AI product performance and client experience closely to gather actionable insights and improve product platform.
• ► Responsible for designing, evangelizing, delivering the first of its kind conversational banking

Sep 2016 – Sep 2017 · 1 yr · Bengaluru, Karnataka, India

• Designed & Architected the AI platform ‘triniti’ from scratch .
• Built the following features for conversational AI :
• 1.Detecting out of domain or adversarial queries in conversational context
• 2.Disambiguating factoids from actionable user queries
• 3.Detecting context change in a conversation
• 4.Identifying compound queries,split and expand to atomic sentences
• 5.Identifying and expanding fragments in user queries
• 6.Handling context expansion over multiple interactions
• 7.Belief tracking dialog system
• 8.Question answering system requiring minimal to no variants in training data
• 9.Improving question answering performance using RL
• 10.Dynamic response reformulation for Greenfield conversations
• 11.Machine comprehension using attentive response reformulation.
• 12.Developing a universal task learning environment
• 13.Handling unplanned conversations in dialog systems
• 14.Developing intelligent memory architecture for driving context driven dialog system
• 15.Developed Dialog manager for handling conversational state , perform query elucidation, probing, slot filling , conversational grounding, Lead-on queries etc .
• 16. Predictive conversations for multi-initiative conversations

Dec 2015 – Aug 2016 · 8 mos

• Led the data science initiatives at IGP, building an AI-propelled ecosystem for Gift discovery platform driven by Advanced Machine learning, Natural language processing, Optimization, Deep learning.
• ► Revolutionary Product discovery platform
• ► Anticipatory Recommendation Framework
• ► Lean, Automated ML modules
• ► Intelligent Search Engine
• ► Evolving Ranking mechanisms
• ► Automatic image tagging & object recognition
• ► Customer behavior modeling: Segmentation & predicting transition within segments

May 2014 – Nov 2015 · 1 yr 6 mos · Innovations - Bangalore

• My role involved developing Machine learning algorithms and their implementation to build data-driven products in a bid to change the holistic experience of consumers ;in the manner that they buy products, perceive a brand, or express their loyalty. Our Machine learning driven product connects many such dots which are ignored in normal brick and mortar stores, using which we can enhance the overall consumer experience providing them highly accurate product recommendations by modelling their preferences, buying behaviour and fashion sense while still keeping the interests of retailer intact by optimising the overall profits.
• Key Highlights:
• 1. Built a Hybrid Product Recommendation system .
• 2. Customer Behaviour Modelling
• 3. Clienteling application
• 4. Social Media Analytics -Text mining and Analysis

Aug 2013 – May 2014 · 9 mos · Bangalore

• AheadYou started product trials recently with one of the largest US universities and is in engagement activation discussions with 10+ institutions including Harvard, Babson, UChicago, Northeastern, University of Texas and one of the largest universities in Asia. Advisors and well wishers include Prof. Bill Sahlman at Harvard Business School, Former President of Babson College (Len Schlesinger), Co-Chair of Rock center at HBS (Tom Eisenmann), Former CEO of Eduventures among others.

Jan 2013 – Aug 2013 · 7 mos · Bengaluru, Karnataka, India

May 2012 – Jan 2013 · 8 mos · Bengaluru Area, India

Feb 2011 – Feb 2012 · 1 yr