Sep 2024 – Present · 1 yr 6 mos · Palo Alto, California, United States · Hybrid

• Series D Investor

Jun 2023 – Present · 2 yrs 9 mos · Stanford, California, United States · On-site

• Generative AI
• "Graphical Models: Uncovering Hidden Relationships in Complex Systems"
• Impact of urbanization, GDP on greenhouse gases

Feb 2023 – Present · 3 yrs 1 mo · Palo Alto, California, United States · On-site

• Applied AI, ML APIs, LLMs
• Creating LLMs for Math visualizations.
• Aug 2025: Inc
• Generative AI, LLMs, Causal Models.
• Autonomous LLMs.
• LLM agents
• Time-Series Foundation Models
• 2023:
• started as an LLC
• working with MIT Horizon as an AI SME

Feb 2023 – Present · 3 yrs 1 mo · Cambridge, Massachusetts, United States

• Curating & Authoring content related to state of the art AI topics such as Generative AI, Large Language Models (LLMs), Deep Learning.
• Consulting vis Pandita AI LLC

Nov 2020 – Oct 2021 · 11 mos · Georgia, United States · Remote

Oct 2018 – Nov 2020 · 2 yrs 1 mo · Greater Boston Area

• Developed perception algorithms that are an ensemble of Deep Learning and Signal Processing methods. These have hit production across all amazon warehouses.
• Developed probabilistic and deep learning methods of object identification on conveyer belts using CV, multimodal Signal and image processing methods such as image deblurring. (Patent)
• Developed an markovian anomaly detection algorithm that can identify anomaly picks by robotic arms.
• (Patent)
• Here is the robot:
• ___________________
• [ ----0 0 0 0 0 0---- ]
• ==@===----===@==

Oct 2017 – Oct 2018 · 1 yr · Seattle, Washington, United States · Hybrid

• Presented relevant ML cases related to generative AI.
• Causal AI algorithms highlighted in spotlight sessions.
• Actively engage with board members and other R&D teams on state of the art industry trends
• collaborate on setting a AI
• roadmap

Mar 2016 – Oct 2018 · 2 yrs 7 mos · Mysuru Area, India

• Worked on AI algorithms.
• Developed novel generative topic modeling algorithm that can disambiguate topics.
• Designed, architected, and implemented an bayesian record linkage system that can adapt to new data with high precision. This beats the standard Fellegi Sunter linkage methods. Contributed to pgmpy - graphical models library
• parsing of anonymous user text to determine hoax data from true data and spatially localizing the activity. This won the Best Presentation Award at ICMLA
• founder & Investor
• Curriculum design for graduate ML, State University of New York
• Developed near optimal routing algorithm for touring solutions in many cities. (Patent pending)
• Publications: "Supervised Negative Binomial Classifier for Probabilistic Record Linkage", http://adsabs.harvard.edu/abs/2019arXiv190803830K

Apr 2015 – Mar 2016 · 11 mos · Greater Boston Area · On-site

• Implemented generative AI algorithms in a cluster computing environment.
• Implemented Probabilistic, Classification and Predictive algorithms.
• Design, build, and maintenance of complex algorithms that perform record de-duplication, entity resolution and probabilistic matching used in the processing of massive heterogeneous data sources to form a high integrity, high quality, clean data asset.
• Design, build and measure data ingestion and integration pipelines for large volumes of temporal data from different sources. Examples include database extracts, application server logs, scanned images, voice recordings, Twitter streams, websites, and health sensor data.
• Spark, Python, Hadoop, MongoDB

Jan 2009 – Jun 2010 · 1 yr 5 mos · Cambridge, USA

• Machine Learning, Language Modeling
• Implemented a novel language model specific to automotive domain using google n-grams, SRI LM toolkit.
• Developed perl scripts to build a comprehensive database consolidating sensor data incoming from various sensors at varying sampling rates and time durations.
• Development and system integration of software and hardware for experimental testing of Real time embedded system.

Sep 2005 – Sep 2007 · 2 yrs · Cambridge, MA · On-site

• Previously BBN Technologies - Speech Group. Worked directly under Prof. John Makhoul and Spyros Matsoukas (part of Alexa founding team).
• Investigated system combination algorithms that combine multiple automatic speech recognition (ASR) system outputs to improve speech quality.
• Programming: Perl, C, Matlab. Platform: Linux.
• Implemented Bayesian Combination (BAYCOM), a system combination method based on Naïve Bayes.
• Developed a novel system combination technique, Confusion Matrix Combination (CMC).
• Parallelized code to run in a cluster computing environment.
• Implemented code that processes large scale data such as 500hrs, 1000hrs of speech data.