V Swapnika

Principal ML Engineer specializing in large-scale recommendation systems, LLM-based ranking, and ML platform architecture serving 100M+ users. Proven record of driving measurable business outcomes and mentoring 200+ ML professionals.

Staff ML Engineer - Indeed - Remote

(2024-03)

Led the technical strategy for the Ranking organization, guiding a team of 11 ML engineers (4 Staff, 5 Senior, 2 Junior) to deliver large-scale recommendation systems that improved Job Applies by ~60%, Apply Rate by ~7%, AUC by ~6%, and served 100M+ users worldwide

• Architected next-gen ranking pipeline (DCN, MMoE, multi-task, LLM) cutting Bad Match Rate ~10% and model size/retraining time ~35%
• Led cross-org evaluation frameworks across Retrieval, Ranking & Monetization, reducing debugging time ~50% and accelerating experimentation
• Productionized Qwen-based ranking system with SFT & GRPO optimization, lifting employer positive outcomes by 48% and improving offline-online metric correlation ~35%

Principal Data Scientist - Chingari - Remote

(2022-07 - 2023-08)

Leading AIML Lab (NLP, Computer Vision)

• Built NSFW detection model combining CLIP transfer learning and object-detection localization, achieving 98% Recall and 87% Precision — outperforming AWS
• Engineered in-app video search engine with auto-complete via Wikipedia scraping and transliteration, aligned to per-user search history
• Designed User Interest Graph Network for personalized feed and search, introducing centrality-based node weighting for authentic user identification

Lead ML Engineer - Dailyhunt (VerSe Innovation) - Bengaluru

(2021-08 - 2022-07)

AI Lab - Leading NLP & Feed Intelligence tracks

• Built multilingual news recommendation system (6 Indic languages) using Graph, Content-based, and CF candidate generators + Two-Tower ranking, lifting CTR 12% and session time 9%
• Developed comment moderation system across 9 Indic languages (2L daily comments) using char-level LM and LSTM, achieving 87% Recall and 92% Precision
• Delivered duplicate article detection for 4L weekly articles using multilingual NER and sentence embeddings, reaching 95% Recall

Data Scientist-II - Amazon - Bengaluru

(2019-08 - 2021-07)

Retail Pricing - Price Error Detection

• Reduced false price-error alerts 81% saving ~64L/year; built Siamese LSTM duplicate-product model at 93% F1 across ~2L consumables ASINs

Data Scientist-II - Amazon - Hyderabad

(2018-05 - 2019-07)

Last Mile Operations

• Saved $135MM worldwide via stop-level service-time prediction; built package-level demand model (22L daily packages) achieving 96% Recall across NA and EU

Data Scientist - Crosslink - Mumbai

(2014-08 - 2016-06)

• Delivered data science solutions in an early-stage environment, building foundational modeling and analytical capabilities

Full Stack Developer Course - Scaler Academy (2022-09 - 2023-06)

PGDBA - Data Science - ISI Kolkata (2016-06 - 2018-05)

B. Tech. - Chemical Engineering - IIT Kharagpur (2010-06 - 2014-05)