Work Experience
Architecting AI agent platforms, enterprise compliance pipelines, and production LLM infrastructure
Current Position
Software Engineer II → GenAI Product Lead · Architect & Technical Lead, Olympus @ Certa
May 2023 - Present- Architected the complete production system for Risk Agent — taking it from early concepts to a reliable, enterprise-grade risk evaluation product.
- Owned end-to-end system architecture: agent orchestration, production infrastructure, service design, multi-tenant vector isolation, and operational reliability.
- Replaced 10–12 hours of manual analyst review per assessment with under 10 minutes of automated execution.
- Evaluated SOC 2, ISO, PCI DSS, HIPAA, and pentest reports across 21 risk domains, 840 controls, and ~2,500 attributes (~8,400 model calls, 210M input tokens at 100-way concurrency).
- Led Azure OpenAI to AWS Bedrock migration analysis, throughput modelling a 10.6x speedup on the heavy control set (12 hours → 68 minutes) and 5–6 concurrent tenants versus 1.
- Established an attribute-level accuracy baseline measured against human ground truth, with root-cause analysis feeding a sub-24-hour evaluation loop.
- Shipped document-type-specialised agents across 13+ compliance document types to eliminate domain conflation.
- Implemented deterministic citation grounding (exact-text extraction and URL injection) ensuring 100% citation coverage for SOC 2 documents and eliminating hallucinated citations.
- Built pre-indexing validation gate cascade (empty text, invalid doc taxonomy, SHA-256 duplicate detection) rejecting bad files before burning inference budget.
- Built V1/V2 dual-version runtime letting both versions serve production traffic side by side, and migrated architecture off serverless Lambda to ECS/Fargate.
- Initiated and architected Olympus as a platform-level engineering initiative, synthesizing patterns from Risk Agent, Contracts Agent, and Data Agent to support all agents once live.
- Designed reusable agent infrastructure that eliminates repetitive foundational work, enabling Product and Engineering teams to ship new agents in weeks instead of months.
- Provided production-grade capabilities from Day One — unified execution, tenant isolation, error taxonomies, and cost/usage telemetry — allowing agent teams to focus entirely on business logic.
- Authored HLD, LLD, and infrastructure design, led architecture review with staff and principal engineers, and secured company-wide sign-off; decomposed delivery into 5 milestones.
- Shipped Milestone 1 Foundations and migrated Risk Agent and Contracts Agent onto the platform, deleting two bespoke LLM layers using a strangler migration pattern with dual-run verification.
- Built a golden-set dual-run regression harness with recorded cassettes and CI gating, executing legacy and platform code paths side by side to prove output-neutrality before cutover.
- Authored core platform principle: documents are tenant assets, not agent assets — keyed by content hash plus tenant, reusable across agents without re-ingestion, with logged access widening.
- Took Contracts Agent from an proof-of-concept idea into a production-ready service in approximately 1.5 months.
- Owned full productionization: microservice architecture, operational infrastructure, load balancing, monitoring, and uptime readiness.
- Deployed directly into the first enterprise client's live legal and contracting workflow.
- Designed MongoDB job lifecycle with idempotency guards, retry counting, a max-3-retry policy with 60s backoff, and an aggregated workflow-status endpoint collapsing 5 frontend polls into 1.
- Implemented structural multi-tenant isolation with custom shard keys in Qdrant, enforced tenant query filters, and per-tenant MongoDB database connections.
- Shipped 5 production agents (clause, metadata, obligations, SLA/KPI, risk summary) and NDA review behind one instrumented entrypoint.
- Built DesignAI solution designer backend, helping pivot Certa toward an AI-first direction that contributed to securing $35M Series B funding; reduced solution-design time by 60% with ~10x performance improvement across Fortune 100 client accounts.
- Transformed a monolithic, non-deterministic AI routing prompt into a deterministic logical routing system with well-defined classification, improving accuracy from 62% to 97%.
- Invented Workflow Context Extractor for GenAI Large Workflow Management (patent filed in the US and India as sole inventor); awarded $1,000 inventor award.
- Co-winner of company-wide Certa-fied Genius Award (August 2025, first-ever cycle); inducted into Hall of Fame.
- Introduced Portkey LLM gateway pattern company-wide with multi-model fallback, retry routing, and cost telemetry.
- Established engineering quality baseline: unit and E2E testing (0 → 75% coverage), CI quality gates, and required PR RCA templates.
- Led Product Data from initial POC to platform vision, delivering coverage across 280+ sanctions lists.
- Defined external-data strategy estimated to drive up to 2.5x revenue uplift.
- Shipped Adverse Media Mapping with a multi-LLM pipeline used in demos with Fortune 100 prospects.
- Built a graph-based entity/UBO intelligence layer linking entities, owners, UBOs, sanctions, and adverse media.
Previous Experience
Product Engineer @ Accubits Technologies Inc
Apr 2022 - Apr 2023- Built OpenMind, a Flask-based microservice for organisational network and employee behaviour analysis.
- Developed a GenAI-powered Chrome extension allowing employees to generate and maintain skill profiles from prompts.
- Led development and deployment across 10+ enterprise-facing projects.
- Built an employee skill inventory platform using ELK and MongoDB that onboarded 750+ employees.
- Built CodeEgo, a multi-tenant Python SAST microservice for automated code security analysis.
Earlier Experience
Research Intern / Project Lead @ Microsoft (Field Data Technologies)
Oct 2021 - Jun 2022Led a team engineering an Unmanned Ground Vehicle (UGV) for non-invasive wildlife-monitoring and endangered-species surveys. Built Python communication libraries for long-range telemetry, command control, and image/sensor data transfer over XBee/ZigBee without internet connectivity. Published research at ICCIS 2022.
Front-end Team Lead @ Kerala Police Cyberdome
Feb 2021 - Sep 2021Led a 4-engineer front-end team for Grapnel, a dark-web crawling, indexing, and analytics platform. Built major Angular front-end components and integrated GraphQL APIs. Contributed to a product that won a ₹10 Lakh national prize and is deployed across law-enforcement agencies. Continues as a volunteer member of the Cyberdome technology group.
Research Intern @ Tata Consultancy Services
Jun 2021 - Sep 2021Developed production-ready Deep Learning model APIs using Flask and TensorFlow. Contributed to an image-based virtual try-on system using GANs and OpenCV.
Founder & Team Lead @ Cisco ThingQbator
Apr 2019 - Mar 2020Pitched BitMedi into the Cisco ThingQbator incubator, was selected into the cohort, and directed a team of 5 through to the investor final, where the project won Best Project and secured seed funding. BitMedi is a secure digital medical-history system built with Django.
Co-Founder @ AgriVision
Jun 2020 - Sep 2020Developed a crop-disease detection model using TensorFlow. Incubated under CIEDC. Won recognitions including YIP State Level and Hash Hackathon Third Prize.