Apr 2024 – Present · 2 yrs 2 mos

• Leading AI-powered document processing at scale — architecting multi-LLM frameworks, agentic pipelines, and large-scale backend systems for enterprise IDP workflows.
• 🔹 Architected a configurable multi-LLM microservice framework (FastAPI) supporting GPT-4 Vision, Gemini, and Claude — enabling dynamic LLM switching via configuration, reducing inference costs significantly at scale.
• 🔹 Designed large-scale async document processing pipelines handling 1,000+ page PDFs — chunking strategies, RabbitMQ orchestration, retry logic, and dead-letter queue handling for fault-tolerant production workflows.
• 🔹 Built taxonomy-driven document classification systems with confidence scoring, rules-based validation, exception routing, and full audit trail / data lineage — enabling compliance-ready, explainable AI decisions.
• 🔹 Implemented Generative AI cost optimization strategies across the Titan IDP platform, reducing LLM inference costs while maintaining accuracy benchmarks.
• 🔹 Acted as primary technical contact for the client: system integration, stakeholder meetings, knowledge transfer, and production go-live support.
• 🔹 Authored HLD/LLD documentation; built gateway and orchestrator services managing product workflows and inter-service communication.

Oct 2021 – Mar 2024 · 2 yrs 5 mos

• Built intelligent document processing microservices and No-Touch Processing pipelines for enterprise insurance clients — delivering automation that directly drove revenue growth.
• 🔹 Designed and implemented a No-Touch Processing (NTP) pipeline with taxonomy-driven classification, confidence scoring, and rules-based validation — achieving 100% confidence document extraction without human intervention
• 🔹 Built document taxonomy schemas for KYC documents, income proofs, bank statements, insurance agreements, and claims — enabling structured, auditable classification and multi-page document assembly.
• 🔹 Developed multiple custom FastAPI and Flask microservices for State-National, a leading US insurance company, handling complex document processing and business automation requirements.
• 🔹 Implemented end-to-end OCR pipelines with multi-page PDF/TIFF support, page-level classification signals, and structured data extraction at scale.
• 🔹 Led and mentored a team of 3 developers across delivery milestones; promoted to Senior Software Engineer in recognition of impact delivered.

Oct 2018 – Sep 2021 · 2 yrs 11 mos

• Joined Infrrd as a founding-stage engineer — built OCR systems, ML training pipelines, and document extraction POCs that became the foundation of Infrrd's core IDP product.
• 🔹 Built multiple POC solutions for automated information extraction from images and scanned documents using OpenCV, Tesseract OCR, and Azure Cognitive Services — pioneering Infrrd's early Document AI capabilities.
• 🔹 Developed custom preprocessing algorithms for OCR text normalization, image enhancement, and PDF/TIFF document handling including large multi-page documents.
• 🔹 Collaborated with the ML team to build end-to-end model training pipelines using YOLO/Darknet and TensorFlow for document object detection and page classification.
• 🔹 Built NER models using spaCy for named entity extraction from unstructured document text — integrated into production classification workflows.
• 🔹 Owned the complete ML lifecycle: dataset generation, data cleaning, model training, evaluation, and inference pipeline deployment into production services.
• 🔹 Used Flask for Python-based microservice development and Elasticsearch as the primary document store and search engine.