Headline

Lead AI/ML Engineer | DS & GenAI | PhonePe-r

About

Lead AI/ML Engineer with 9.5 years building and leading production ML systems across consumer fintech, e- commerce, and algorithmic trading. Combines deep technical expertise (distributed ML pipelines, XGBoost/LightGBM, transformer architectures, LLM agents) with product-oriented delivery: designed A/B testing frameworks (100K+ users), built personalization stacks (retrieval → ranking → rerank), and productionized GenAI (RAG, multi- agent, LLM-as-Judge). Leadership track record: grew AI team 1→7, directed quarterly ML roadmaps with cross-functional stakeholders, mentored 3+ data scientists. Quantified impact: $5.6M+ annualized revenue, 15–20% cross-sell lift, 13% risk model improvement, 11% propensity uptake increase.

Location

Dubai

AI stack

antigravity

codex

opencode

Work experience

Lead AI Research Engineer at Rapid Global Growth, May 2025 - Jun 2026 · 1 yr 2 mos, Dubai, United Arab Emirates, - A/B Testing & Dynamic Pricing: Built foundational A/B framework (103K visitors, Bonferroni-corrected) validating segment-level price elasticity. Deployed 10-model LightGBM ensemble with per-channel isotonic calibration across 8 price bands, natively handling class imbalance. 11.08% revenue lift → $5.6M–$6.1M annualized. - Cross-Sell Personalization & Ranking: Designed end-to-end retrieval and ranking system for checkout-page accessory recommendations (candidate generation, LightGBM ranker, EV-optimized reranking with margin normalization and diversity constraints), replacing fixed-profitability carousel. Achieved 15% accessory attach rate lift, generating $60K/year incremental revenue. - EV-Optimized Ranker: LightGBM predicting calibrated p(click) & p(purchase|click) per price band (isotonic regression). Margin-rate-normalized EV scoring with category/price-tier diversity constraints. 15–20% accessory attach lift → $80K/yr incremental. - Multi-Agent Trading Platform: News-triggered system (planning → tool dispatch → prediction) for intraday direction. 12 parallel structured-data fetchers, hybrid LSH-embedding deduplication, custom RoBERTa NER (92.8% F1), <10ms retrieval. LLM-as-Judge eval (1,200+ trades, CAPM-adjusted returns) isolating skill from luck. Pre-fetched structured data improved accuracy 66%→76%. - RAG Candidate Support: RAG pipeline auto-resolving inbound queries via email. RAGAS-evaluated chunking/retrieval; HITL escalation for low-confidence responses → 24 hrs/week manual effort saved with accuracy/compliance preserved.

Lead Data Scientist at PhonePe, Apr 2023 - May 2025 · 2 yrs 2 mos, Bengaluru, Karnataka, India, • Architected and deployed a Gen-AI solution processing 1B+ SMS messages daily, reducing processing time from weeks to 15 minutes through distributed computing optimization and achieving 85%+ entity recogni- tion accuracy, enabling targeted marketing campaigns that improved mutual fund CTR by 15% and reduced portfolio risk by 5%. • Designed and deployed a production feature store for the Merchant Lending division, reducing feature com- putation time from 3 days to 2 hours and enabling real-time decisioning across 100M+ transactions daily, supporting credit risk, propensity modeling, and collections use cases. • Developed and deployed a credit risk scorecard processing 1B+ UPI transactions bi-weekly, improving portfolio risk prediction by 13% and reducing delinquent customer targeting errors by 20% through advanced feature engineering and XGBoost ensemble methods, directly impacting loan approval accuracy and collections effi- ciency. • Built an automated model monitoring framework tracking statistical drift (PSI/CSI) for production models and reducing model degradation incidents through proactive performance management. • Led development of a loan propensity model that increased uptake by 11% through customer segmentation and targeting optimization, partnering with Product and Risk teams to design A/B tests and translate model insights into acquisition strategies, while mentoring a senior data scientist on advanced feature engineering techniques. • Directed quarterly ML roadmap planning for Merchant Lending in partnership with Heads of Lending, Product, Risk, and Collections, prioritizing initiatives across credit scoring, customer targeting and collections while providing technical mentorship to 3 data scientists.

Lead Data Scientist at ZestMoney, Apr 2022 - Apr 2023 · 1 yr 1 mo, Bengaluru, Karnataka, India, • Developed and deployed a credit risk scorecard using XGBoost on transaction, credit bureau, and behav ioral data, improving loan allocation efficiency by 17% and reducing rejections by 4% through real-time risk assessment. • Built a fraud detection system using document similarity and pattern recognition to identify fabricated bank statements across thousands of loan applications, reducing first-payment defaults by 11% through automated verification integrated into the production approval pipeline. • Designed an automated document processing service using computer vision and OCR, processing 10K-25K Aadhaar cards daily and saving 1M INR per month in operational costs while ensuring regulatory compliance for customer data protection. • Developed a customer lifetime value optimization framework comprising 5 models i.e loan propensity, amount prediction, and repayment forecasting, implementing a simulation-based approach to prioritize customer out reach that reduced communication costs by 10% while maintaining conversion rates, demonstrating ROI-driven ML application. • Grew ZestMoney’s AI team from 1 to 7 engineers over 3.5 years, establishing production ML standards for model deployment, monitoring, and governance while maintaining hands-on contributions to high-impact projects across fraud detection, credit scoring, and customer analytics.

Artificial Intelligence Engineer - II at ZestMoney, Oct 2020 - Mar 2022 · 1 yr 6 mos, Bangalore Urban, Karnataka, India

Artificial Intelligence Engineer - I at ZestMoney, Sep 2019 - Oct 2020 · 1 yr 2 mos, Bangalore Urban, Karnataka, India

Junior Associate Consultant at Jaitra Inc, Jul 2017 - Sep 2019 · 2 yrs 3 mos, Bangalore, karnataka, • Built pharmaceutical document auditing tools using NLP and semantic similarity to automate compliance verification across 1K-2K regulatory documents, reducing manual review time and improving accuracy. • Developed a medical search engine with OCR and named entity recognition, providing clinical trial officers with unified access to drug specifications across multiple databases, eliminating time-consuming manual searches. • Designed document management microservices with full-text search and version control capabilities for pharmaceutical workflows.

Intern at JUSPAY, May 2017 - Jun 2017 · 2 mos, Bangalore, India, A pre-placement offer(PPO), I was introduced to UPI. Technical skills acquainted include Haskell and Purescript.

Education

Bachelor's degree, Computer Science, Svce

SSC, HSC, Kendriya Vidyalaya No. 1, Jalahalli West, Bangalore - 15