Chhaya Tundwal

OurFirm.ai builds large-scale legal intelligence systems by transforming raw court documents into structured, searchable data. • Built a large-scale AI document processing pipeline for 10M+ legal opinions, transforming court documents into structured JSON and production-ready HTML. • Designed a page number and citation recovery system that preserves reporter citations, footnotes, and metadata across heterogeneous sources including HTML, plain text, and PDFs. • Developed Python data processing pipelines to normalize legal documents and recover missing structural elements such as page markers, paragraph boundaries, and citations. • Wrote complex PostgreSQL queries across the CourtListener dataset to analyze opinion clusters, citations, and metadata relationships across millions of records. • Implemented automated validation checks (12+ rules) to detect rendering issues such as missing page markers, broken footnotes, and dropped text before documents are published. • Optimized LLM extraction pipelines using Gemini (Vertex AI) to process millions of documents while minimizing token cost and reprocessing. • Managed supporting AWS infrastructure (RDS, EC2, IAM) for large-scale document processing pipelines and database environments. Scale: 10M+ legal documents processed, multi-million record datasets.

built a fully functional System Design Tutor web application that guides learners through multi-level system design challenges using dynamic, AI-generated level plans. I engineered the platform end-to-end in Next.js/TypeScript, designed a robust state management layer to handle multi-step workflows, and implemented reliable level-evaluation logic that advances only when the AI explicitly signals readiness.

Automated prompt-driven code documentation pipeline using Python, Gemini Pro, Cursor and SQL, cutting manual documentation effort 70% and migration time 50%. Designed ETL workflows with n8n + LangChain, aggregating 5K+ monthly customer reviews and did sentiment analysis. Created a prompt-engineering workflow using Cursor and Gemini Pro to auto-generate comprehensive documentation for an AngularJS codebase, significantly reducing manual writing effort. Led the migration of an AngularJS application to React.js using prompt engineering to streamline code translation.

Automated ETL workflows with Python & SQL to integrate Google Analytics & Salesforce into unified reporting views. Collaborated with faculty and marketing teams to translate business requirements into dashboards (Tableau, Looker). Improved campaign conversion rates by 12% through actionable insights derived from centralized data pipelines. Built Looker Studio dashboards integrating Google Search Console, LinkedIn, and Salesforce data for real-time campaign insights. Improved PMP website performance from 39 → 90 on PageSpeed Insights, increasing engagement and SEO visibility. Implemented A/B testing & analytics pipelines for digital campaigns, boosting CTR and lead conversion.

Designed Python pipeline to transform LinkedIn data into structured formats, boosting lead qualification accuracy by 80%. Integrated data into NoSQL stores (MongoDB, DynamoDB) and streaming agents for scalable enrichment and retrieval. I worked on simulating AI-powered B2B sales pipelines using large language models (LLMs) integrated with LangChain and Prediction Guard’s secure GenAI platform. My responsibilities included designing and implementing agentic AI modules to automate key sales tasks such as lead research, qualification. Evaluate AI agent performance against human benchmarks across metrics like cost, speed, and accuracy. Additionally, contributed to building a simulation environment that helped the company benchmark AI autonomy levels, optimize internal sales workflows, and create a case study demonstrating responsible AI adoption for enterprise clients.

Built and scaled data pipelines using Kafka (Pub/Sub), DynamoDB Streams, and Oracle PL/SQL, transporting and validating 10M+ financial transactions/day with automated governance checks that prevented $12.5M in potential losses. Migrated 100+ legacy workflows from COBOL and on-prem systems to AWS Step Functions, Batch, and EKS, enabling event-driven automation and reducing manual intervention. Designed REST APIs and backend services in Java (Spring Boot) and AWS Lambda to securely access customer and transaction data for analytics, audit, and compliance. Engineered ETL and data validation layers in Python, implementing schema checks, lineage documentation, and Pytest-based quality gates for reliable ingestion and transformation. Optimized analytical workloads by designing Snowflake and BigQuery schemas, reducing query runtimes by 40% and improving data discoverability across teams. Standardized CI/CD pipelines with Terraform, Jenkins, and Docker, cutting deployment times by 40% and improving developer velocity by 60%. Implemented Kafka event publishers streaming 100K+ daily account events, improving downstream observability and near-real-time reporting. Monitored production workloads with Datadog, Splunk, and CloudWatch, maintaining 99.9% uptime and <5-minute MTTR during on-call rotations. Partnered with compliance and business teams to modernize Account Opening and EFT workflows, increasing scalability and delivering $10M+ annual savings.

Developed a robust automation pipeline in Python and PL/SQL stored procedures to process 1M+ transactions/day, performing CRUD operations on Oracle databases and triggering alerts when business thresholds were exceeded.

Developed a Decision Support System (DSS) for Multimodal Freight Transportation Planning for Food Corporation of India(FCI) Kerala Region, which helped reduce the cost by $2 million and transporting time of grains by months. Deployed the web app, database and Nginx using Docker Compose, GitHub Actions and scripting. Social Impact of the project was that it reduced the food wastage as the time and weight of the freight was distributed according to the capacity and requirement in the region.