Multi-agent AI & grounded retrieval
7 LangGraph pipelines, prompt contracts, Qdrant hybrid retrieval, and citation grounding that reject unsupported claims.
Production AI,built to verify.
I’m Avinash Sinha, Founding Engineer at Bull AI and former CTO. I build multi-agent research engines, document AI pipelines, and cloud infrastructure where every answer traces back to source data.
Systems in production · Since 2021
From exchange filings and OCR tables to citation-locked research agents and 45 AWS service stacks—I build software where model output, data lineage, and infrastructure operations stay auditable.
↓Own the full path.Keep the evidence attached.Design for safe recovery.
I’m a Founding Engineer at Bull AI in Mumbai, where I own system design, full-stack engineering, and production operations across customer-facing applications, multi-agent research pipelines, financial document ingestion, and AWS infrastructure. Before Bull AI, I spent four years as CTO at BigPluto Technologies, leading a team of 10+ engineers across real-time RFID tracking for Maruti Suzuki (10,000+ users and assets), NFC campus platforms, digital-twin inspection workflows, and GPU video pipelines. Having carried both executive delivery risk and hands-on on-call responsibility, I optimize for verifiable outputs, reversible cutovers, and systems another engineer can debug at 2 a.m.
See core capabilitiesCapabilities & Architecture
Document AI & financial data ingestion
Layout-aware Paddle OCR, chart-to-table extraction, Rust/Python GPU workers, and NSE/BSE filing deduplication.
MCP servers, Go APIs & full-stack apps
48-tool MCP finance server, OAuth 2.0/PKCE gateway, Next.js research workspaces, and auditable Razorpay billing.
Cloud infrastructure & event-driven systems
45-root GCP-to-AWS Terraform migration, Kafka/RabbitMQ telemetry for 10,000+ assets, and GPU media pipelines.
Engineering leadership & governed AI workflows
Former CTO for 10+ engineers, 1,500+ auto-reconciled knowledge records, 180 runbooks, and 92 architecture decisions.
Built for production constraints
Tools change every quarter; sound boundaries do not. I group my work by the operational problems a system has to survive.
↓Ingest
Capture NSE/BSE announcements, XBRL results, transcripts, and SAST filings; deduplicate bytes and resolve canonical company and investor identities.
Structure
Parse XBRL directly into financial facts, and run layout-aware Paddle OCR, chart-to-table recognition, and LLM validation on complex PDFs.
Synthesize
Route jobs across 7 LangGraph multi-agent pipelines, combining Qdrant vector evidence with PostgreSQL financial facts under versioned prompt contracts.
Deliver
Enforce citation and schema gates, run point-in-time backtests, and serve traceable research via Next.js SSR, Go APIs, and a 48-tool MCP gateway.
Where I add leverage
on day one.
Production AI Systems
Multi-agent LangGraph pipelines, prompt contracts, RAG over financial filings, and 48-tool MCP servers that ground model output in verifiable data.
Document & Data Pipelines
Layout-aware Paddle OCR, chart-to-table extraction, Rust GPU workers, and idempotent NSE/BSE ingestion with canonical entity resolution.
Cloud & Distributed Backends
Go and Python services, 45-stack AWS Terraform migrations, and Kafka/RabbitMQ event architectures built for sub-second updates and safe rollbacks.
Technical Leadership
Zero-to-one founding execution combined with prior CTO experience leading 10+ engineers, 92 ADRs, 180 runbooks, and governed AI coding workflows.
Experience & Technical Stack
Five years of shipping systems.
Bull AI
Founding Engineer · Applied AI, Backend & Cloud
Own system design, full-stack engineering, and production operations across 7 LangGraph research pipelines, a 48-tool OAuth/PKCE MCP finance server, Go APIs, Next.js SSR workspaces, Rust/Python Paddle OCR workers, NSE/BSE filing ingestion (5,300+ canonical investor IDs), ~45 AWS Terraform service roots, KMS PII encryption, Razorpay + GST billing, and 1,500+ governed knowledge records.
Python · Go · TypeScript · Rust · Next.js · LangGraph · Qdrant · AWS · TerraformBigPluto
Chief Technology Officer · Led 10+ Engineers
Directed technology strategy, microservices architecture, AWS infrastructure, and CI/CD delivery across four production platforms: Lighthouse (real-time RFID tracking for 10,000+ users and assets at Maruti Suzuki via Kafka and RabbitMQ with sub-second updates), Learntank (integrated NFC attendance, fleet, and cashless payments), Homecertify (digital-twin property inspection workflows), and Thrillfactory (GPU-accelerated FFmpeg pipeline handling 500+ sessions daily).
Kafka · RabbitMQ · Microservices · AWS · GPU/FFmpeg · NFC · RFIDCloudwood & Opecise
General Manager - Technology at Cloudwood Entertainment (2021–2022): designed and implemented an OTT content delivery and user-management platform using React, GraphQL, and Node.js. Software Development Engineer at Opecise (2021): built Next.js SSR SaaS modules and reusable Node.js backend APIs.
Software engineering since 2021Explore case studiesLanguages: Python, Go, TypeScript, JavaScript, Rust, SQL, Bash, HCL. AI & Data: LangGraph, RAG, MCP, Qdrant, pgvector, Paddle OCR, PostgreSQL/Supabase, Kafka, RabbitMQ, Redis/Valkey. Cloud: AWS (ECS/Fargate, EC2/GPU, SQS, EventBridge, Step Functions, KMS, Bedrock), GCP, Terraform, Docker.
Building a critical system?
Let’s talk architecture.
I’m best suited for Founding Engineer, Staff/Senior Backend & Applied AI Engineer, and AI Platform roles where a team needs end-to-end ownership across multi-agent systems, document/data pipelines, Go/Python backends, and AWS infrastructure.
Email avinashsinha0226@gmail.com