Vipila Bathvar
I’m an AI/ML Engineer with 8+ years of experience building and deploying production-grade ML systems, especially in RAG, retrieval evaluation, and low-latency inference services.
About
I’m an AI/ML Engineer with 8+ years of experience building and deploying production-grade ML systems, especially in RAG, retrieval evaluation, and low-latency inference services. I’ve led end-to-end work from model development and MLOps automation to monitoring and safe releases using MLflow, CI/CD gates, and rollback strategies—delivering measurable improvements to reliability, retrieval precision, and time-to-answer.
I enjoy turning complex requirements into practical, SLA-driven AI deliverables across financial and insurance use cases. My background also includes OCR-to-structured pipelines, streaming/data orchestration, and building robust Python APIs (FastAPI/Django) with strong observability and security practices, so teams can confidently operate ML in production with clear KPIs like citation coverage, cost-per-query, and p95/p99 latency.
- 9 years' experience
- Joined September 2026
Industry experience
- Financial Services
- Software & Internet
Work experience· 4 roles
Architected a production RAG system on AWS Bedrock + LangChain with Pinecone/FAISS, delivering grounded answers with configurable guardrails and retrieval evaluators. Deployed FastAPI inference on GKE with HPA and resource tuning to maintain predictable p95 latency. Implemented MLflow experiments and Model Registry with stage-gated promotions and automated rollback on regressions using pre/post-deploy quality checks. Built an OCR-to-JSON pipeline with OpenCV + Tesseract plus Pydantic validation, reducing manual data entry and downstream errors. Automated retraining via SageMaker Pipelines triggered by drift/performance thresholds, and added offline RAG evaluation suites with CI quality gates for prompts/indices. Built Iceberg + PySpark lakehouse patterns federated to Snowflake with schema evolution and data quality checks, and tuned hybrid retrieval (BM25 + dense) for compliance-sensitive queries. Instrumented OpenTelemetry for latency/error hotspot diagnosis and secured endpoints with
Built Django REST APIs for core banking workflows with RBAC, audit trails, and idempotency for reliable transaction handling. Optimized ORM/SQL and implemented caching to improve throughput and reduce p95 latency during peak hours on AWS. Split monolithic endpoints into microservices deployed to Kubernetes using blue/green releases. Implemented async tasks for reconciliations, notifications, and reporting. Built ETL and data migrations to Snowflake/PostgreSQL with validations and monitoring hooks for data quality. Automated CI/CD with Jenkins/Git and added unit/integration/contract tests to prevent regressions. Improved observability with CloudWatch/Prometheus and implemented API usage analytics, quotas, and client-key visibility. Conducted load testing and supported DB tuning/caching decisions, while enforcing secure coding with dependency scans and secrets rotation.
Shipped Django/Flask services and REST/SOAP integrations to unify partner systems under consistent contracts. Containerized legacy apps and standardized Helm charts for Kubernetes deployments. Built serverless jobs with Lambda and EventBridge for bursty workloads to reduce baseline cost. Implemented Glue/Step Functions ingestion/transform pipelines with schema checks and alerting, and established testing baselines with pre-merge quality gates. Hardened public endpoints using API Gateway throttles and WAF rules, and centralized secrets/configs across environments. Optimized JSON serialization/pagination for payload size and latency, added distributed tracing with correlation IDs, and exported curated datasets to Snowflake/BigQuery for downstream analytics. Also contributed to GUI development and generated graphical reports using Python visualization libraries.
Hire Vipila Bathvar today
Post a job and invite Vipila, or contact them directly. No hidden commission.