Software Engineer · Remote, India

Pranav Krishna
Danda

Software engineer building AI agents and backend systems that work in the real world — real users, real data, real pressure.

View My Work Get In Touch Download Resume
Scroll
Recently Shipped → MergeSignal (pre-merge risk intelligence) · Writing at pranavkrishna.hashnode.dev

About Me

I like problems that look messy at first.

I’m a software engineer based in India with 2.5 years of applied experience building AI agents and distributed backends. Most recently at Oovacha, where systems I worked on run across 27 production deployments in a regulated environment. Before that, I built ML pipelines, RAG chatbots, and Django APIs at Mushroom Solutions.

I learn fastest by shipping, and I’m happiest on teams that move fast and debug out loud. B.Tech in Computer Science & Engineering (AI & ML) from VVIT.

Open to full-time roles
8 Production AI Agents Shipped at Oovacha
27 Production Deployments Live in regulated environments
78.6K + Records Processed In production pipelines
15 Skill Packs Authored On the Claude Agent SDK

Experience

Software Engineer

Oovacha

Sep 2025 — Present
  • Architected a 6-phase distributed pipeline (SQS + ECS Fargate, S3-synced state, Postgres JSONB checkpointing) turning long-form domain PDFs into validated persist SQL — multi-hour jobs with checkpoint-resume, DLQ, retry classification, and fail-closed validation on every write.
  • Designed and own end-to-end a multi-endpoint onboarding API backend with a monotonic 5-stage state machine, single-transaction multi-table writes, tenant-scoped RBAC (cross-tenant probes collapse to 404), and AWS Step Function orchestration — every write audit-logged for regulatory compliance.
  • Wired multimodal LLM components into the pipeline (vision + text via Amazon Bedrock + Claude Agent SDK) — authored 15 modular agent capability packs and shipped 8 production agents, each fenced by deterministic Python guards and a golden-regression eval that gates every prompt change before merge.
  • Built a queue-driven validation pipeline processing 78.6K+ records against 77 configurable rules on ECS + SQS + Aurora + S3 + Cognito; integrated custom MCP servers so non-technical users query the domain database in natural language.
  • Live across 27 clinical trial sites in production.
PythonFastAPIAWSPostgreSQLAmazon BedrockClaude Agent SDKMCPSQSStep Functions

Software Engineer Intern

Oovacha

Mar 2025 — Aug 2025
  • Ramped up on clinical data systems and AWS infrastructure while contributing to the initial EDC validation pipeline
  • Built tooling and test coverage for agent workflows; promoted to full-time Software Engineer in Sep 2025
PythonFastAPIAWSPostgreSQL

Software Engineer Intern

Mushroom Solutions

Mar 2024 — Mar 2025
  • 12-month internship — promoted from Full-stack Developer Intern to ML Engineer Intern in Sep 2024.
  • Trained deep learning classification models hitting 97.8% accuracy on 15,400+ records — automated 17 workflows, eliminating 22 hrs/week of manual processing
  • Built an intelligent RAG chatbot combining Azure AI Search + FAISS with Redis caching, serving 5,000 queries/day
  • Shipped Django REST APIs with role-based access control, serving ML inference across 4 internal teams
  • Covered agent tool execution and edge cases with end-to-end test automation using Behave BDD
TensorFlowKerasFAISSDjangoRedisAzure AI Search

Featured Projects

Multimodal Manifest Pipeline

6-phase SQS-driven ECS pipeline turning long-form domain PDFs + N vendor spec PDFs into a validated data-collection manifest and idempotent persist SQL. Combines Claude Opus vision (multi-page table extraction at 230 DPI) with deterministic pdftotext parsing for row-precision data. Fail-closed validation on every LLM write; signed-override path for high-trust data.

PythonAmazon BedrockClaude Agent SDKSQSECS FargateS3

Proprietary — no public repo

Study Onboarding Backend

Multi-endpoint API backend with a monotonic 5-stage state machine, single-transaction 8-table skeleton writer, tenant-scoped RBAC (cross-tenant probes collapse to 404), and AWS Step Function orchestration for downstream provisioning. Every write structured-audit-logged for regulatory compliance.

PythonFlaskPostgreSQLStep FunctionsLambdaSecrets Manager

Proprietary — no public repo

Clinical AI Agent Framework

3 autonomous agents with tool execution, error recovery, and MCP-based natural language querying of production PostgreSQL databases.

PythonClaude APIMCPFastAPI

Proprietary — no public repo

EDC Validation Engine

Message-driven 3-stage pipeline processing 78.6K+ clinical records against 77 configurable rules. SQS-backed for message durability.

PythonSQSPostgreSQLAWS

Proprietary — no public repo

RAG Search Platform

Hybrid vector + keyword search behind Django APIs. 5,000 queries/day, sub-second latency across 10,000+ document corpus with Redis caching.

FAISSAzure AI SearchRedisDjango

Proprietary — no public repo

Skills & Technologies

AI & Agents

Amazon Bedrock Claude Agent SDK MCP Protocol Multimodal LLMs Autonomous Agents RAG FAISS Azure AI Search LlamaIndex TensorFlow Keras

Backend & APIs

FastAPI Flask REST APIs Pydantic SQLAlchemy Django

Cloud & Infrastructure

AWS ECS Fargate AWS Lambda Step Functions SQS Aurora PostgreSQL S3 Cognito Secrets Manager CloudWatch Docker

Databases & Messaging

PostgreSQL Redis RabbitMQ

Frontend

TypeScript Next.js React

Testing

Behave BDD Test Automation

Writing

I'm documenting what I build — agent architectures, backend design decisions, and lessons from production AI systems.

Published Jul 2026

Building a Thinking-First Multi-Agent Room (Not Another Fake Agent Demo)

Multi-AgentArchitectureClaude
Read post
Published May 2026

I Gave an AI Agent Access to a Production Database. Here's What Actually Happened.

MCPAI AgentsProduction
Read post
Published Sep 2026

Building MergeSignal: Pre-Merge Intelligence for Git

GitDeveloper ToolsStatic Analysis
Read post
Follow on Hashnode →

Let's Work Together

I'm actively looking for full-time Software Engineer roles focused on AI systems and backend infrastructure. Open to roles across India and remote opportunities globally.

Response time: usually within 24 hours.

Email pranavkrishna317@gmail.com LinkedIn Connect with me GitHub View my code
Available for full-time roles · Based in Andhra Pradesh, India