Consulting on end-to-end architecture, roadmap planning, setting up development processes and team building for their SaaS product, chat based AI analyst called BYOB where you can integrate various data sources and analyse your data using LLMs.Product Link: https://www.akaike.ai/byob-by-akaike

I had privilige of working across AI/ML, Backend and Infra during my time of 6 years at Healthify as detailed below:AI/ML Product- Built pre ChatGPT versions of Ria (chat based AI coach), developing most of the ML and backend infrastructure with a small team.- Built core components of Ria, data access layer, rule evaluation engine, config management and content generation, as a platform to power realtime insights across the product.- Led the integration of ChatGPT levaraging most of the existing infrastructure to help launch new version of Healthify app.- Designed metabolic scores for easy interpretation of continuous glucose monitor readings working with resident physicians.Machine Learning & MLOps- Rearchitected the diet plan generation to be more personalised improving NPS for the whole product from -10 to 30.- Built time series classification and glucose response prediction models with sparse CGM and smart watch data.- Built models for photo based food recognition (Snap), from data collection to multi GPU model fine-tuning.- Built custom reusable MLOps pipeline on AWS Sagemaker for continuous training and deployments.Backend- Built multiple extensively reusable and scalable systems to power personalised and actionable insights across the app.- Led the backend development for multiple product launches of Ria, Smart Plans, Pro Plans (CGM) and Snap.Infra / DevOps- Led the migration of backend services from stand-alone instances to Amazon ECS (container orchestration).- Single handedly managed up-time of both backend and data infrastructure during pandemic times handling twice the traffic.- Led successful migration to terraform and multiple efforts for cost optimisation, backend migrations, database upgrades and migrations and many critical high-traffic events.- Built the infrastructure team with the right systems, skills and processes for monitoring distributed service deployments and incident management.

Built a smart, fully automated and responsive team and lead management system for SMBs to help optimize operations with the mobile first approach. (www.betyphon.com)

>> Built easy-to-use web-based data analysis and visualization tools: ------ Mosaic - to track and detect data inflections in real-time across hundreds of tickers and multiple datasets; and filter by correlations, growth rates and sectors.------ Plotter - to visualise, analyse and compare hundreds of fields across financial, web & social, market, ESG or custom data sets; and parts of other products across the platform.>> Developed algorithms to extract sentiment from earnings transcripts, financial documents and domain-specific tweets and articles.>> Developed linear models to reduce variety of data sets into single time series and designed condensed metrics to help compare across sectors and companies.>> Built automated systems to collect data from variety of public and private data providers.>> Developed a high-level interface for easy customization of data pipeline and for fine-tuning the models while decreasing the computation run-time by 3X using data cache.>> Developed ad-hoc methods for outlier detection, to handle missing/unreported data and to identify and process time series data from structured data sets with minimal supervision.

>> Built a Face detection platform customized to improve customer experience in retail stores as an improvement to existing loyalty programs and to aid real time analytics.>> Designed a cluster based estimation and ranking algorithm to determine a fair auction price of a player in the context of IPL, including data gathering from web.