Aditya.

Designed and shipped a graph-based agent orchestration system powering Wordsworth's core product — multi-provider scraping, structured extraction, and real-time Redis streaming, generating production-ready webpages at scale.

Productionised a large-scale document-intelligence system over 500+ documents (300 — 500 pages each). OpenSearch + a reasoning-based retrieval layer. 88% accuracy on internal benchmarks.

Engineered a real-time event-capture pipeline (Pub/Sub + Apache Beam) tracking user behaviour across sessions — 100GB+/day through BigQuery, handling 1,200+ req/sec at peak.

Shipped a CRO agent combining engagement and conversion signals: 25 — 150% conversion uplift across key e-commerce funnels. Established eval and benchmarking pipelines for PDP optimisation, CRO agents, and QA. Architected GCP infra for 10+ services: Cloud Run, Cloud Build, a Playwright/Selenium grid, and queue workers.

Joined the big-data team. Built cross-departmental Spark/Hadoop ETL pipelines consolidating multiple business units for firm-wide cash forecasting — replacing manual Excel workflows with automated nightly batch jobs and reducing ETL runtimes by 40%.

Developed statistical and Transformer-based cash-forecasting models on the consolidated platform — 85% accuracy, supporting $38M in liquidity decisions. Optimised internal treasury pipelines via partition sizing, executor configs, and shuffle tuning; HDFS small-files compaction lifted throughput 3× while cutting runtime 50%.

Designed Elasticsearch indices for low-latency P&L search. Built a Text-to-SQL system using LLMs for ad-hoc natural-language querying of financial data. Side-stint on blockchain: smart contracts for a bond-issuance platform. Fast Track promotion plus multiple performance awards.