Dec 2021 – Jan 2024 · 2 yrs 1 mo · Bangalore Urban, Karnataka, India

• Highlights
• Led Copilot team to develop generative AI feature suite, transforming the way recruiters and hiring managers collaborate with each other and engage with candidates.
• Launched the first AI text generation use case in the product, using OpenAI models. Implemented streaming to reduce effective latencies, and caching to reduce OpenAI request costs
• Led Skill extraction from candidate profiles to enable better skill based hiring
• Launched and led the annotations for evaluation datasets that helped me to recommend feature engineering improvements for job-candidate recommendation models - leading to 2% improvement in recommendations for some minority classes in our data
• Launched internal resume parser, extracting personal contact details and educational qualifications, and set up pipelines to track metrics and latencies on production data and to compare new models.
• Resolved customer issues, critical technical bugs and on-call issues AI Platform team.
• Directly involved in ML Hiring pipeline and interviewing candidates, on-boarding, guiding and mentoring interns and new graduates amongst the top talent pool, resolving blockers and ensuring project success.
• Researched and presented learning sessions on new ML solutions
• Awards
• September, 2023 Wall of Fame for leading Generative AI development charter
• Most Innovative Hack Award (from among ~80 teams). Led a team of two to re-imagine dashboards for hiring analytics. Talked to various stakeholders, understood the current pain points, developed and demoed an ML based solution within a short time frame.This has since been launched to completely revamp dashboard experience for our users.

Mar 2020 – Dec 2021 · 1 yr 9 mos · Bangalore Urban, Karnataka, India

• Highlights
• Envisioned and developed an in-house product to derive insights from earning call transcripts for ~500 public companies. Conducted extensive research of third party tools available to create Product Feature Documents and Architecture Design documents to convert the above project into a competing product at scale.
• Introduced a pipeline for information retrieval (company names and company descriptions) from unstructured PDF documents using GPT-2, DistilBERT (for text classification), and text similarity using transformer embeddings. This can be utilized by analysts to create reports, saving 60-70% manual reading effort.
• Launched a (Named Entity Recognition) NER API that trains models and auto-annotates documents, saving ~500 man-hours of annotation effort. This was employed to train and version models to extract product opportunities from meeting notes/call transcripts and to automatically tag documents and from client emails for automatic ticket creation.
• Devised interactive dashboards to a) analyze client emails in shared mailboxes, b) present insights from earning call transcripts for ~500 public companies.
• Used Graph Variational AutoEncoder (GVAE) to calculate node similarity in a knowledge network graph of potential clients.
• Awards
• 2nd place in company wide hackathon to utilize data on upcoming startups - from Glassdoor, Pitchbook, Crunchbase to support research in investment teams
• 4th place in company wide Hackathon to derive insights from public and proprietary data to assist in Investment Strategies for ESG outcomes/Growth Equity.

Jun 2019 – Mar 2020 · 9 mos · Greater Bengaluru Area

• Released a fully functional POC that performed super resolution on videos on edge devices - Desktop and Android (Qualcomm DSP), which got the first client - a top video streaming company.
• This POC proposed to improve the OTT media users’ experience and reduce the bandwidth usage by at least 30%.
• Researched current architectures, performed end-to-end ML modeling, created a POC Android app to integrate the computer vision models into a product pipeline.

Apr 2018 – May 2019 · 1 yr 1 mo

• Business Case study for Sales Prediction and Marketing Mix: Trained a custom Neural Network in Tensorflow to predict sales based upon investment in different promotional channels and lagged sales. The model took synergic effects of various channels into account which led to 71% reduction in MSE error in comparison to the previously used regression models.
• Responsible for development of web apps for Business Rule Execution and ETL for automating Sales Crediting and Incentive Compensation processes for a multinational Pharmaceutical client - powering the analytics behind the flagship product SalesIQ TM, v.4.0 through v.4.5
• This app was built with modular, meta-data driven, multi-threaded, scalable technologies, integrating Salesforce.com, AWS (S3, Redshift, Athena) in an Agile Development environment.
• Developed a web app for converting text to speech, introducing additional features such as parsing text from single or multiple files (txt, doc, docx, ppt, pdf) to convert it into downloadable mp3 speech files. The service is now being used internally by teams in the organization.

Jun 2017 – Mar 2018 · 9 mos

Jun 2016 – Jul 2016 · 1 mo · Chandigarh, Chandigarh, India

• LMS| Moodle- Development and Administration| PHP| Javascript| Web Development
• Full stack developer for an online learning platform offering courses in entrepreneurship and related fields.
• Assisted in conducting the research on e-learning methodologies to select the right tools for developing an e-learning management system.
• Worked on designing and developing required plugins for Moodle LMS.

Jun 2014 – Jun 2014 · 0 mo · Chandigarh, Chandigarh, India

• I handled client projects to optimize their websites and blogs to get more hits and hence more revenue. I worked on both on-page and off-page search engine optimisation techniques and was able to significantly improve the rankings of the clients' sites and blogs.

Jun 2009 – Jun 2009 · 0 mo · Chandigarh, India · On-site

• Selected to represent school for a Times of India Workshop/internship. Brainstormed, planned, organized the articles to be printed alongside some of the best minds in journalism, wrote a few editorial pieces, edited some others to finally publish a student edition newspaper that was circulated over the city.