Hi, my name is Ankit Josh.

I build efficient systems.

I'm a Software Engineer at DataCore, building production AI systems for enterprise support.

I work with backend engineering, distributed systems, retrieval, and agentic AI.

Python · C++ · Go · RAG · LLMs · Agents

Selected Work

DataCore Software · Internal Platform

Atlas

I architect and co-lead development of Atlas, an internal AI support and diagnostics platform. It ingests tickets, knowledge, support bundles, and live cluster telemetry to help support engineers triage issues and investigate likely root causes.

The platform combines LLMs, RAG, MCP-based tools, asynchronous workloads, and product-specific pipelines behind a shared multi-product architecture.

Highlights

  • 13 active support agents
  • 40 cases handled in a week
  • 200K+ embeddings
  • 6× faster reindexing
  • <2ms query latency

Technologies

  • Python
  • FastAPI
  • Celery
  • PostgreSQL
  • Redis
  • Elasticsearch
  • Milvus
  • LLMs / MCP

Atlas is an internal system, so implementation details and source code are not publicly available.

Selected Projects

Systems Project

PyKeyDB

A Redis-inspired, thread-safe in-memory key-value database built from scratch with transactions, write-ahead logging, persistence, and asynchronous networking.

Designed around a layered Protocol → Session → Execution Engine architecture with concurrent access and crash recovery.

View code

Technologies

  • Python
  • asyncio
  • Threading
  • WAL
  • Transactions
PyKeyDB

AI Project

VisionText

A multi-modal image retrieval system supporting text-to-image and image-to-image search using CLIP-based embeddings and Qdrant.

Includes hybrid retrieval and ranking using Reciprocal Rank Fusion, an async FastAPI backend, batch processing, persistent model caching, and Dockerized deployment.

View code

Technologies

  • Python
  • FastAPI
  • CLIP / SigLIP
  • Qdrant
  • Docker
VisionText multi-modal image search

What I enjoy building

Backend systems — APIs, concurrency, data stores, asynchronous workloads, and distributed services.

AI systems — retrieval, RAG, LLM orchestration, agents, tool use, and production AI infrastructure.

Systems from scratch — I like understanding how things work by building them myself, from key-value stores to retrieval pipelines.

Let's connect

Interested in backend engineering, AI systems, distributed systems, or building something interesting?

Email me