Currently building Aryra — an AI-powered assistant combining multimodal analysis, agentic reasoning, semantic search, sandboxed code execution, and inline content generation in a single streaming interface.

Aryra is a production AI platform featuring:

• Custom ReAct Agent: 33-tool agent loop (no LangChain/LangGraph), real-time SSE streaming of reasoning and Plotly charts
• Provider Abstraction: Single env var switches entire cloud stack — AWS (S3 + Textract + Bedrock + Transcribe) or GCP (GCS + Cloud Vision + Vertex + Google Transcribe), with per-service overrides. LLM layer supports Anthropic, AWS Bedrock, Google Vertex AI (Claude + Gemini), OpenAI, and Ollama under a unified interface
• Knowledge Bank & HNSW: pgvector semantic search with HNSW indexing for sub-millisecond nearest-neighbor lookup, HyDE query expansion, auto-caching of every Q&A pair as embeddings for progressive knowledge enrichment
• Multimodal Asset Analysis: Image OCR (Textract or Cloud Vision), PDF page-level extraction with per-page embeddings, Excel chunked processing, audio transcription (AWS or Google Cloud Speech), video analysis via native Bedrock video blocks (s3Location, Nova Pro)
• Inline Image & Video Generation: Abstract GenerationProvider (Bedrock, OpenAI DALL-E, Google Imagen) — generate images and videos directly in chat, streamed as file uploads
• NL2SQL: Schema-aware natural language to SQL with automatic schema discovery, TTL-cached metadata, and 20+ injection pattern blocks
• File Management & Deduplication: SHA-256 hash-based deduplication across S3/GCS, proxied downloads, semantic file search, storage lifecycle tiering
• Redis Caching: Abstract CacheBackend (Redis + NullCacheBackend fallback), TTL-aware schema deduplication, semantic query caching
• Self-Hosted Observability: OpenTelemetry (FastAPI + asyncpg instrumented) → Grafana Alloy → Loki (logs) + Tempo (traces) + Prometheus (metrics) + Grafana dashboards
• Session Sharing & Forking: Fork any conversation from any message, token-based sharing, fork chain traversal, bookmarks, real-time SSE notifications
• IAM-Style RBAC: JSON policy documents, deny-wins evaluation, role hierarchy, wildcard resource matching across 8 resource groups
• Sandboxed Python: subprocess isolation, custom MetaPathFinder blocking 18 dangerous imports, inline Plotly charts, Excel/PDF/PPTX/CSV export
• Integrations: Gmail OAuth (12 tools, multi-account), SearXNG + Crawl4AI web search, Google SSO, persistent per-user memory

Tech Stack: Python · FastAPI · React 19 · TanStack Start · React Native/Expo · PostgreSQL/pgvector · Redis · AWS Bedrock · GCP Vertex AI · Docker · OpenTelemetry · Grafana

Founder at Aryra (Jan 2026 - Present)

• Built full agentic AI platform with custom ReAct agent (33 tools), SSE streaming, and multi-LLM routing across Anthropic, AWS Bedrock, Google Vertex AI (Claude + Gemini), OpenAI, and Ollama
• Provider abstraction layer — single env var switches entire cloud stack (AWS or GCP) with per-service overrides
• Self-hosted observability: OTel → Grafana Alloy → Loki + Tempo + Prometheus + Grafana
• React 19 / TanStack Start SSR frontend with pnpm monorepo, Turborepo, and React Native/Expo mobile app

Senior Software Consultant at Piping Rock India (Mar 2025 - Present)

• Built hybrid OCR + Vision asset analysis API with AWS Textract + Claude Sonnet 4.5 and Amazon Titan V2 embeddings
• Designed LangGraph ReAct agent with 16 tools and Nova Embed V2 (3072-dim) multimodal embedding pipeline
• Deployed FastAPI on AWS Lambda via Docker + SAM with 99-file test suite and full audit logging

Senior Software Engineer at Headout (Mar 2024 - Oct 2024)

• Scaled review collection system to 1.5M users with 11.4% increase in submissions
• Designed WhatsApp Business API integration achieving 7.3% collection rate increase
• Led full-stack project with Kotlin microservices, Redis caching, and Mixpanel analytics

Software Engineer at Cimpress (Dec 2023 - Mar 2025)

• Optimized PostgreSQL queries with 1.4x to 27x performance improvements
• Led UI component development epic for Order Management System
• Designed traceability system with DocumentDB for distributed systems

Software Developer at Piping Rock (Jan 2022 - Dec 2023)

• Built Warehouse Management System on AWS reducing operational costs by 33%
• Developed custom ORM and WebSocket microservices
• Improved application load times by 20% with database connection pooling

📄 Download Latest Resume (PDF)

Auto-generated from the resume repo on every push to main.

Languages:

Frontend:

Backend:

Databases:

AWS:

GCP:

AI Providers:

DevOps & Observability:

• Scaling Aryra to a wider user base at aryra.in
• Expanding provider coverage and fine-tuning multi-LLM routing strategies
• Growing the knowledge bank with richer graph interlinking and smarter context enrichment