Senior Software Engineer, AI at Xoriant (2026-02 – Present)
- Re-architected Finastra's legacy RAG log analyzer - leading a 2-engineer team to build a zero external dependency DeepAgents ReAct system, reducing investigation from days to minutes across GB-scale logs.
- Designed high-performance ingestion and agentic flows using row-group pruning and tool-call deduplication, reducing redundant LLM context and lowering costs by ~45%.
- Engineered a DuckDB-based analytics layer with Parquet files stored in OneDrive, enabling sub-second queries across millions of log records and eliminating the need and cost of a dedicated external database.
- Deployed the app on Azure App Service; built layered defense-in-depth AI security - prompt injection detection, SQL sandboxing, access isolation, and tool-failure recovery for regulated financial data with goal and trajectory based agentic evaluations.
- Delivered live forensic platform demonstrations to Finastra engineering teams - achieving ~75% investigation depth on the first query for real-world payment failures.
- Built Impact Analyzer Copilot using GitHub Copilot SDK and Finastra MCP capabilities with SSE streaming, serving 200+ users. Extended Copilot's capabilities with custom cloud agents, skills, hooks, and instructions.
- Extended Operator Assist with Oracle 26ai (HNSW vector store, conversation manager) for financial institutions; identified and fixed a critical RAG grounding issue related to retrieved context.
Engineer, AI at Altimetrik (2025-08 – 2026-02)
- Delivered SchemaFlow AI for Lululemon - led HLD and built a LangGraph multi-agent platform automating cross-system schema management, cutting cycle time by ~50%.
- Built a repo-aware code integration agent that materialized agent-generated code into production-ready GitHub PRs with automated merge/conflict resolution, structure-aware file placement, and branch management via GitHub MCP.
- Designed a canonical knowledge layer with hybrid RAG (ChromaDB vector + Neo4j graph) and incremental-ingestion adapters, achieving ~70% accuracy on requirement-to-implementation impact analysis.
- Productionized the platform on AWS with LLMOps using LangSmith tracing, runtime analytics via AgentCore.
Associate Engineer, AI at Altimetrik (2024-01 – 2025-07)
- Built a secure, scalable GraphQL backend with GenAI integration for Lululemon Enterprise Asset Management; delivered SSO, API protection, and assisted in CI/CD via GitLab CI + AWS ECR.
- Designed a PostgreSQL schema compatible to a text-to-SQL agent and KPI-driven analytical workflows.
- Led HLD/LLD for GenNexus multi-agent system with Jira/Confluence/Datadog integration, cutting manual triage effort by 65%.
- Delivered a multimodal AI Fashion Designer platform with runtime trend forecasting, reducing design cycle time by ~35%.
- Shipped internal productivity copilots: HR chatbot (40% faster ticket resolution) and SEO content generator (55% faster content production).
- Drove GenAI adoption through structured enablement sessions, mentoring teams, and AI capability hiring.
Associate Engineer, Product & Platform Engineering at Altimetrik (2022-06 – 2024-01)
- Built and optimized microservices & features for Samsung India e-commerce & retail platforms; integrated payment gateways, third-party APIs, and messaging systems.
- Optimized payment/refund workflows - reduced refund issues by 70% and improved API response time by 30%; hardened security with OAuth 2.0 and AES-256 encryption.
Application Development Associate at Accenture (2022-02 – 2022-06)
- Supported Converged Cloud platform operations, ensuring platform stability.
