March 2025 – Present · India

Engineering leadership

• Lead the India ML/AI engineering team, set technical direction, mentor engineers, and review architecture and implementation.
• Own delivery and production readiness across enterprise AI platforms, from problem formulation and PoC to evaluation, deployment, and operations.

Commercial Pricing & ML Decision Intelligence

• Architected a production pricing decision framework combining transaction-level feature engineering, comparable-price logic, recency and transaction weighting, conversion-rate adjustments, ML recommendations, and multi-layer commercial guardrails.
• Operationalized reusable data and model pipelines through Kedro and Argo, with lineage, regression validation, and memory/runtime optimization.
• Designed an agentic “Explain the Price” system that connects model outputs, costs, discounts, order-size adjustments, and active experiments, with follow-up questions on comparable deals and win likelihood.
• Architected a pricing engine covering ~$1.6B in trailing-12-month revenue, built on ~$8.8B in historical transactions; controlled experiments contributed $13.5M+ in estimated annualized margin impact across eight tests.

Customer Attrition & Enterprise Churn Modeling

• Designed a leakage-safe customer attrition model using 31 behavioral, volume, product-mix, quote, and lead-time inputs over four-quarter observation windows to predict 50%+ next-quarter volume decline.
• Converted churn probabilities into revenue and volume at risk, identifying approximately 2,430 significant-reduction customers per quarter representing roughly 243M T12M lbs, and translated the scores into intervention priorities and controlled validation plans.
• Built an enterprise portfolio churn framework combining financial, contractual, organizational, people, and engagement signals into entity-level predictors of relationship risk.

Experimentation & Economic Signal Intelligence

• Designed and evaluated controlled pricing experiments across material/form adders, large-order discounts, recommendation methods, and adoption scenarios, reconciling win-rate movement with gross-margin economics through sensitivity and break-even analysis.
• Built a signal-engineering pipeline across 52 external indicators, mapping signals to end markets, applying STL seasonal adjustment and PPO transformations, and validating relationships against end-market-specific volume.

Quality Intelligence, GraphRAG & Applied AI

• Lead the design of a structural-intelligence platform for large codebases, combining deterministic program analysis, graph-based reasoning, evidence-gated decisions, explainability, and large-scale evaluation.
• Architected and benchmarked an enterprise GraphRAG system combining entity-relationship graphs, semantic traversal, and vector retrieval; achieved 89% multi-hop fact recovery versus 53% for a leading enterprise benchmark and 2.9× faster average inference across 22 ground-truth questions in six categories.
• Built a formal retrieval and LLM evaluation framework that separates retrieval quality from answer quality and includes adversarial and out-of-scope cases.
• Improved internal GraphRAG evaluation quality from 43.1% to 54.6%, with zero confirmed hallucinations in the reported evaluation after retrieval, traversal, and grounding refinements.
• Led architecture for a production content-intelligence platform across a multi-million-article corpus, combining hybrid retrieval, ranking, recommendations, entity intelligence, external trends, and geospatial signals; reduced manual curation by 50%+.
• Designed LLM-assisted tagging, drafting support, retrieval augmentation, and distribution workflows using BigQuery, Cloud Spanner, and Vertex AI, with prompt/version control, tracing, observability, and regression checks.

Generative Vision & Multimodal AI

• Re-engineered node-based generative workflows into modular Python services using Stable Diffusion XL, LoRA, IP-Adapter, SAM-HQ, and GroundingDINO for image-conditioned generation, segmentation, and controlled multimodal workflows.
• Added stage-level tracing, artifact validation, model/config versioning, deterministic controls, and regression checks to support reproducibility and production readiness.

June 2022 – February 2025 · India / UK

Voice AI, Language Technology & Digital Commerce

• Co-founded and engineered a mobile-to-mobile, real-time voice translation platform with approximately 0.45-second latency, supporting 3,000+ bilingual calls during pilot programs.
• Built multilingual translation systems that achieved 20% higher fluency across Hindi, Tamil, Bengali, and Gujarati against baseline benchmarks.
• Led ONDC and Bhashini integrations and voice-assisted merchant onboarding workflows, reducing listing time by 40% and supporting 500+ sellers.
• Deployed multilingual AI for Parliamentary Archives covering 1M+ records and enterprise voice workflows processing 50K+ minutes per month, contributing approximately $1.5M in annual savings.

BFSI Decision Intelligence for a Confidential Client

• Led the architecture of an AI-powered credit-underwriting platform processing 7M+ loan applications annually, shortening approval cycles by 45% and reducing processing errors by 20%.
• Developed ensemble risk models using 10+ years of loan history, macroeconomic indicators, and behavioral signals, improving risk-prediction accuracy by 30%.
• Designed a GenAI-assisted counterfactual decisioning approach that helped eligible rejected applicants move toward approvable profiles, increasing approval rates by 18% and unlocking $20M+ in estimated annual revenue.
• Reduced default rates by 2.5 percentage points and underwriting costs by 10%, generating approximately $3.5M in annual savings, while unstructured-data analysis improved fraud-detection accuracy by 35%.
• Built an LLM- and speech-analysis-driven customer reactivation framework that increased conversion from approximately 3% to 19.8% and delivered $15M+ in annual savings.
• Enabled Kafka-based recovery and lead-generation analytics that improved recovery efficiency by 25%, lead-to-conversion rates by 40%, and customer engagement by 30%.
• Delivered a multilingual customer and representative assistant handling 1.5M+ daily interactions, with OCR-based document extraction at 98% accuracy; reduced query-resolution time by 40%, improved customer satisfaction by 25%, and increased cross-channel adoption by 50%.
• Added computer-vision document checks that reduced onboarding fraud cases by 20% and deployed Docker-, Kafka-, and AWS-based infrastructure capable of handling 25× peak traffic, maintaining 99.99% uptime while lowering infrastructure cost by 35%.

Healthcare, Enterprise AI & Applied Research

• Delivered a voice-enabled health agent for elderly care, supporting 2,000+ seniors and reducing manual caregiver workload by 30%.
• Developed predictive flu-outbreak models with The Weather Company by combining climate and health signals, achieving approximately 80% accuracy with up to a two-week lead time.
• Led enterprise voice-bot and agent deployments that reduced inbound query-handling costs by 35%, and explored multimodal integrations with expressive 3D avatars.

August 2021 – May 2022 · China / India

• Developed and scaled multilingual conversational AI for English, Spanish, and Portuguese, handling 1M+ monthly interactions with 99.9% uptime.
• Integrated real-time computer vision for automated document verification, achieving 95% accuracy and reducing manual review by 30%.
• Connected NLP and vision services with CRM workflows, reducing query-resolution time by 25%.

January 2019 – August 2021 · Hyderabad, India

Research collaboration with L. V. Prasad Eye Institute

• Collaborated with clinicians to develop CNN-based keratoconus progression models using tomography data from 1,200+ subjects.
• Supported progression prediction up to approximately 500 days in advance through medical-image modeling, longitudinal validation, demographic analysis, and explainable ML.

Conversational AI & Recommendation

• Designed an ensemble intent-classification system achieving 93.5% accuracy with sub-0.65-second latency.
• Developed a hybrid content-based and collaborative recommendation system with automated retraining and multi-channel support.

• AI Consultant, SDLC Corp — May 2017 – January 2019
• Software Engineer, SDLC Corp — June 2015 – April 2017
• Programmer Analyst, Cognizant — August 2014 – June 2015