Headline

ML/DS @ Roku

About

Staff Data Scientist with 8+ years of experience in applied ML, optimisation, NLP, deep leaning and causal inference. Proven track record in building large-scale credit risk, payments, and automation systems at CRED and Samsung. Adept at converting complex business problems into scalable, production-ready solutions.

Location

Bengaluru

AI stack

anthropic apis

claude code

claude

codex

openai

Work experience

Senior Software Engineer, Machine learning at Roku, Jul 2026 - Present · 4 mos, Bengaluru, Karnataka, India, ML and DS for Roku Pay (RPAY); Payment routing, Subscriptions, Fraud and others

Staff Data Scientist at CRED, Dec 2024 - Jun 2026 · 1 yr 7 mos, Handled ML/DS initiatives for Payments, Lending and Data enrichments; Offer management and Pricing for Lending - CRED worked as a LSP and had integration with more than 6 partners which have 10+ partner policies actively running - Architected pricing and offer allocation for all CRED members using optimization with dynamic offer visibility rules to manage risk and also keeping an eye for incremental disbursals - Improved ~3% incremental disbursals with ₹1.5 Cr+/month in incremental revenue with zero OPS dependency Tech stack: Optimization (LP/MILP/Non-linear; Gurobi, DEAP), Spark, Databricks, AWS Batch Alternate Bureau for realtime underwriting - Mined financial signals from SMS, Email, UPI, and NPCI data to build alternate bureau that provides >90% accurate information (matches with bureau) for ~60% of available users. Used that to cut overall NPA by ~8% - TAT (Turn around time) for getting the information reduced from ~30-60 days to <1 day Tech Stack: SQS, NER, SpaCy, transformers, Docker

Lead Data Scientist at CRED, Dec 2022 - Dec 2024 · 2 yrs 1 mo, Handled ML/DS for Bill and rent payments; BBPS Automation - Architected and built auto-scraping and event-driven reminder pipelines basis of which CRED can show bills and remind them automatically without user intervention; - Handled scale of 60M events/day in a realtime process along with proper monitoring for data correctness and latency - ~60% of all payments happens through accounts created through our system. Generating ~₹70L+/month MDR revenue and our reminder logic helps in cutting fetch costs from ₹1 Cr to ₹55L/month with no degradation of user payments Tech Stack: Python, Spacy, CatBoost, AWS SQS, Lambda, Airflow, Docker, LLM Agents

Senior Data Scientist at CRED, Dec 2021 - Nov 2022 · 1 yr

Lead Engineer at Samsung Electronics, Mar 2019 - Dec 2019 · 10 mos, Bengaluru Area, India, - Handling schedule recommendation task for different DA devices end to end from data preparation to modelling Tools used- Spark for Data preparation, Ensemble Methods for modelling of users

Senior Software Engineer at Samsung Electronics, Jul 2017 - Feb 2019 · 1 yr 8 mos, Bengaluru Area, India, - Working in IoT domain as a Machine learning model developer for different predictive requirements - Developing skills in Spark along with other Big Data tools that is required for data preparation and modelling

Education

Master's degree, Statistics, Indian Statistical Institute, Kolkata

Bachelor's degree, Statistics, University of Calcutta