Passionate about building industry-first AI products from scratch.
Earlier, built an AI tutor@Bytelearn as part of the founding team. Ex-Chief Scientist at the startup ShieldSquare acquired by Radware (now Radware bot manager). Later, I led the R&D to launch Airtel’s first app-based voice assistant.
At Airtel xLabs (newly formed R&D wing of telecom giant Airtel):
– Responsible for R&D on speech and NLP for the next billion. Building deep learning-based solutions to address unique India-specific challenges: Speech Recognition and NLP in low resource (vernacular), low bandwidth (e.g., 2G), code-mixed (e.g., Hindi-English mixed) settings. Deployed ASR system for Airtel’s mobile app
– Devised solutions using transfer learning, custom augmentations and data enrichment to address the unique setting, resulting in significant improvements in accuracy (e.g., ~6x reduction in WER within vertical, ~10% improvement in WER at low bandwidth)
At ShieldSquare (acquired by Radware):
– Initiated and led R&D on ML for bot detection as Chief Scientist. Hired and built a top-tier applied ML R&D team from scratch. Radware acquired ShieldSquare for its advanced ML capabilities in bot detection ( https://www.radware.com/newsevents/pressreleases/2019/radware-to-acquire-shieldsquare )
– Led R&D on novel ML-based solutions which improved accuracy by ~20%. 3 patent applications filed (novel unsupervised, Multi-armed Bandit and LSTM-based models for bot detection)
– Led full-stack R&D of industry-first bot detection solutions in new domains, from conception to deployment (led release of 2 products: Javascript-based Clean Analytics and Ad Fraud mitigation)
– Cross-functional contributions on high priority projects: partner/M&A-related presentations, white papers and blog articles for marketing, support and custom R&D for high priority enterprise customers, analyst presentations, development and devops
– Responsible for scalability and architecture as the incoming load increased from 100s of millions of queries per month to 6 billion queries per month (12x) in ~6 months. After achieving horizontal scalability in architecture, started focusing on DS/ML R&D
Before joining ShieldSquare, completed my PhD work in Distributed Information Retrieval at Dept. of CSE, IIT Madras in 2015 (joined for MS and updated to PhD).
Fun fact: implemented and trained a deep neural network in pure SQL using BigQuery: https://www.linkedin.com/pulse/deep-neural-network-implemented-pure-sql-over-harisankar-haridas-phd/
Goodreads profile: https://www.goodreads.com/user/show/5106448-harisankar