Forward Deployed AI EngineerData EngineerSolutions Architect
I’m currently a Forward Deployed AI Engineer at TechTorch, where I build AI solutions for private equity-owned companies and modern businesses worldwide.
Over 10+ years, I’ve worked across analytics and data engineering. The journey took me from wearing multiple hats at startups to building large-scale ad targeting pipelines at Amazon Devices, then owning the AI data platform behind Amazon’s AI Seller Assistant. Today, I bring that foundation into building AI solutions that fit how customers actually work.
01 / SELECTED WORK
Data infrastructure, AI products, and decision systems.
Here’s what I owned and who it served.
Founding AI Data Engineer / Data Platform Engineer
I owned the data foundation behind Amazon’s AI Seller Assistant: streaming and analytics, knowledge and RAG pipelines, and a self-service evaluation hub—from architecture and ingestion to tooling and production reliability.
Built the streaming-to-warehouse foundation for Seller Assistant analytics and leadership reporting.
Capture high-volume LLM telemetry across requests, retrievals, evaluations, feedback, and cost—while keeping events consistent as they arrived at different times and in different formats.
Designed and built a medallion data platform processing hundreds of thousands of LLM trace signals per second. Owned ingestion, enterprise data modeling, and shared analytical datasets.
Implemented idempotent, replay-safe processing, automated quality checks, failure recovery, and historical reprocessing. Defined fact and dimension models with incremental processing, late-arriving data handling, backfills, and SCD Type 2 history.
Science, Product, Business, and partner teams used the curated datasets for evaluation, benchmarking, adoption, latency, cost, and product performance. Shared metrics supported weekly and monthly business reviews through senior leadership.
Built knowledge ingestion and RAG pipelines that connected seller guidance with seller-specific context.
Prepare policies, help content, and operational guidance across source formats, regional policies, languages, and marketplaces for reliable retrieval.
Owned source ingestion, document normalization, and retrieval-context preparation. Prepared policy and website content for knowledge-base indexing and RAG pipelines.
Combined shared seller knowledge with seller-specific data so the assistant could draw on relevant guidance and the seller’s situation when answering questions about their business.
Built the full-stack workflow for configuring reviews, collecting annotations, and analyzing assistant quality.
Connect product teams defining quality, auditors reviewing responses, and science and analytics teams interpreting the results in one repeatable workflow.
Owned the full-stack experience and data processing for sampling configuration, rubric setup, multilingual audit generation, and annotation transformation.
Product managers configured what to sample and how to assess it. Human and model-based review workflows returned annotations that became structured, analytics-ready datasets, giving downstream teams consistent data to assess quality and inform improvements.
Built targeting pipelines that turned terabytes of traffic into modeling features and real-time ad publishing inputs.
At flat advertising budget.
Turn large-scale device traffic into useful targeting candidates and dependable features for modeling and real-time ad publishing.
Owned end-to-end targeting pipelines, feature preparation, and release guardrails. Prepared query-to-product candidates and features covering impressions, clicks, click-through rate, organic rank, and eligibility.
Added quality checks, alerting, and feedback loops to improve feature quality and release safety. The work increased attributed units by 27% and attributed revenue by 23% at a flat advertising budget.
Built credit-risk models and decision systems to support lending growth and identify repayment risk.
Across the combined credit-risk and fraud-detection work.
Help finance and product teams balance loan growth with portfolio quality through better credit assessment, fraud detection, and early risk signals.
As an early data science hire, I partnered with the CFO and led four analysts. My ownership covered credit modeling, risk analytics, analytical infrastructure, and team leadership.
Connected model development to finance and product decisions. The combined credit-risk and fraud-detection work reduced delinquencies by 58% and increased loan conversions by 37%.
Built an AWS-native platform that audited Seller Data Protection Policy compliance across organizations and generated violation tickets.
Designed a last-impression attribution model and Tableau serving layer to help teams evaluate search, campaigns, and content decisions across 8+ marketplaces.
For the full work history and technical background, see my resume.
02 / MY STORY
University of California, San Diego
BS Computer ScienceGuru Gobind Singh Indraprastha University
Cars24 gave me my startup DNA. AI Seller Assistant showed me what it takes to build for millions.
At Cars24, I was an early data science hire working directly with the CFO and leading a small team. The work crossed analytics, fraud, credit risk, and infrastructure. It taught me to step beyond a job title, understand the business, and take responsibility for getting something useful built.
Amazon Devices deepened my engineering experience. Building ad targeting pipelines across global marketplaces made reliability, data quality, and operational discipline part of how I work.
Becoming the founding data engineer for AI Seller Assistant brought those experiences together. I helped build the data foundation of a large-scale AI product, with responsibility spanning knowledge, evaluation, high-volume streaming, and organization-wide analytics. It changed how I think about ownership: the platform has to serve the product, its users, and the teams improving it.
That’s what draws me to forward deployed AI engineering today. I get to work close to customers, bring a broad technical perspective, and stay hands-on as a solution takes shape.
03 / WHERE I CAN HELP
From finding useful AI opportunities to building the systems behind them, I work with teams to turn business needs into practical solutions.
Open to conversations about roles, consulting collaborations, and pro-bono projects.
TechTorch