Aishwarya Pasnur

Engineer with 3.5+ years building distributed backend systems and applied ML (SageMaker pipelines for demand forecasting), and most recently production LLM/agentic AI engineering with LangGraph. Led a Terraform IaC migration across 15 AWS microservices and architected a three-tier caching layer cutting latency to under 5 milliseconds. Shipped an enterprise AI-assisted invoice-audit platform using LangGraph orchestration, LLM-based extraction and compliance checks, human-in-the-loop governance, and a hash-chained audit trail.

Comfortable owning a system end to end, from the distributed backend up through the AI layer running on it.

AI Engineer at CogniWissen Private Limited (2026-07 – 2026-10)

• Designed multilingual voice AI agents (LiveKit, Sarvam, Cartesia, GPT-4o) with a latency-tiered trigger system, routing simple intents to regex and GPT-4o-mini and reserving full GPT-4o for complex cases
• Built an enterprise AI-assisted accounts-payable audit platform: designed LangGraph orchestration that ingests invoices from email (MS Graph/Gmail) or manual upload, extracts structured fields via LLM, fetches PO/receipt data from ERPNext and QuickBooks Online, and runs deterministic plus LLM-based 3-way-match checks
• Implemented a human-in-the-loop (HITL) review workflow using LangGraph's interrupt/resume pattern and a certified model-governance layer (versioned prompts, pinned model bindings via a LiteLLM proxy) to prevent uncontrolled model drift
• Designed a tamper-evident, SHA-256 hash-chained audit-event system in PostgreSQL and integrated Microsoft Presidio for PII redaction, giving every AI and human decision a compliance-ready audit trail
• Technologies: Python, FastAPI, LangGraph, LiteLLM, PostgreSQL, Microsoft Presidio, Docker, LiveKit, Deepgram, Cartesia, Sarvam AI

Software Development Engineer (SDE-II) at WinGage Technologies (2025-05 – 2026-07)

• Built and scaled 15 Python, Go, and FastAPI microservices for a multi-tenant WhatsApp-commerce platform, integrating WhatsApp Cloud API, Meta Flows, and Razorpay payments with real-time WebSockets and JWT authentication
• Designed the cart-management service (layered FastAPI/SQLAlchemy, optimistic concurrency via version fields, Valkey-backed session resolution), with full Playwright end-to-end coverage
• Designed a Go-based campaign/messaging service and a configurable promotions and discount engine, with Razorpay webhook verification and order reconciliation
• Architected a centralized three-tier Valkey caching layer across config, cart, session, and tenant data, cutting warm-cache reads to under 5 milliseconds versus a 91 millisecond baseline
• Led a Terraform IaC migration across 15 microservices on AWS (dev/preprod/prod), with Bitbucket Pipelines OIDC auth and a scheduled nightly-shutdown Lambda for non-prod cost savings
• Built a multi-tenant web hosting service handling DNS provisioning, SSL via ACM, and CloudFront CDN, and documented a bastion-host procedure for migrating a private RDS instance with no public access
• Applied linear regression to model order volume and campaign response from historical data, and built a 3-stage SageMaker pipeline (Feature Store preprocessing, 20x hyperparameter tuning with SHAP, batch inference) for a demand-forecasting model
• Technologies: Python, Go, FastAPI, SQLAlchemy, PostgreSQL, Valkey, AWS (Lambda, SageMaker, S3, SQS, RDS, CloudFront, ACM), Terraform, Docker, Playwright

Software Developer (SDE-I) at WinGage Technologies (2024-02 – 2025-04)

• Developed backend features and REST APIs in Python and FastAPI for the WhatsApp-commerce platform, implementing webhook-based event handling across AWS SNS, SQS, and SSM
• Containerized services with Docker and validated API contracts using Postman
• Technologies: Python, FastAPI, Docker, Postman, AWS (SNS, SQS, SSM)

Python Developer at Medeva.io (formerly Juxt Smartmandate Analytical Solutions) (2023-03 – 2024-02)

• Filtered and profiled clinical and medication datasets using Python, pandas, and NumPy, then built a Django web application with Matplotlib charts to visualize and present the results
• Technologies: Python, Django, pandas, NumPy, Matplotlib

B.Tech in Electrical, Electronics and Communications Engineering – Dr. Babasaheb Ambedkar Technological University (2022)