Sep 2020 – Present · 5 yrs 8 mos

• Enabling organizations productize their ideas with AI.

Jan 2019 – Sep 2021 · 2 yrs 8 mos

• Responsible for formulating and delivering AI technology road map for the organization.
• Leading a small and efficient team of ML engineers, conceived, designed, and implemented NLP products resulting in 30% improvement in operational efficiency of the organization.
• Main products delivered:
• Legal text generation (NLG) system for generating paragraphs of patent text from an input of few lines using transformer based seq-seq models: GPT2 and T5. Conceived novel methods for generating training data for text generation.
• Custom semantic search algorithm for patent pdf documents using BERT and XLNet involving training a language model from scratch, optimizations to the algorithms and novel ways of generating training data.
• Realtime and offline auto complete using n-gram language model integrated with Microsoft word.
• Text to figure conversion product using information extraction algorithms for patent documents including custom entity detection, event detection and relations between entities and events using dependency parsing, constituency parsing, semantic role labelling etc.
• Filed one patent for the above work.

Jan 2011 – Dec 2018 · 7 yrs 11 mos

• Image processing and Computer vision: Developed cloud and edge based image detection and classification algorithms using different deep learning algorithms for smart home and smart city applications.
• Audio analytics: Developed sound detection and classification solution for detecting custom sounds using custom deep CNN for smart home applications.
• Predictive Analytics: Developed machine learning algorithms for interactive and predictive planning of hundreds of wireless devices in large industrial plants, buildings and homes. Led development of indoor localization platform for large buildings and industrial plants.
• 8 granted patents and two publications.

Nov 2008 – May 2011 · 2 yrs 6 mos

• Primary areas of research: information theory, stochastic optimization and wireless communication systems. Developed stochastic algorithms to for data transmission with Shannon secrecy.
• Published in IEEE Signals, Systems and Computers Conference ASILOMAR

Aug 2005 – Sep 2008 · 3 yrs 1 mo

• Implemented various signal processing algorithms including DLS-MIDI music synthesis standard and AMR wide band audio decoder on Qualcomm DSP. Led design and implementation of different gaming audio post-processing features of 3-D positioning, reverberation, Doppler, chorus and coning.
• Awarded Qualcomm trade secret for method on optimal memory structuring of MIDI state variables for efficient access between hardware and DSP.