Backend Infrastructure & AI Systems Engineer
Hi, I'm Vimal Sahani — IT undergrad at IIIT Bhopal. I design high-throughput backend engines (Django, FastAPI), scalable microservices, and reliable LLM evaluation pipelines. Core open-source contributor to TensorFlow, Django, Celery, and Hugging Face.
Vimal Sahani
IIIT Bhopal (B.Tech IT '27)
GitHub Pro 🎓19+
Merged PRs
5+
Systems Built
GitHub Achievements
Featured Engineering Projects
High-concurrency distributed systems, LLM evaluation platforms, and backend engines.
AI Agent Evaluation & Reliability Platform
Enterprise-grade benchmarking platform to evaluate LLM agents on accuracy, latency, tool invocation fidelity, and hallucination drift. Features automated CI evaluation pipelines, Dockerized runtime, Redis cache, and asynchronous Celery workers.
socialFeed: High-Throughput Feed Engine
Distributed feed generation engine built with Django REST Framework and FastAPI. Engineered with hybrid fan-out on write/read, keyset cursor pagination for zero offset skips, Min-Heap ranking, Redis caching, and documented with 20 Architectural Decision Records (ADRs).
DevPilot: RAG Codebase Intelligence
Production-grade developer intelligence platform for GitHub codebases. Integrates AST parsing, RAG-powered vector search across repositories, automated issue debugging, and contextual pull request reviews with LLMs.
FluxMesh: Distributed Job Orchestrator
Distributed task and job orchestration platform with intelligent scheduling, exponential backoff retries, multi-tier task priorities, worker node heartbeat coordination, and fault-tolerant state recovery.
EdgeFlow: High-Performance API & Model Gateway
Low-latency API reverse-proxy gateway engineered for intelligent request routing, token bucket rate limiting, distributed caching, circuit breaker retries, and automated service & LLM model failover across backend nodes.
Open Source Contributions (19+ Merged PRs)
Real-world production patches merged into industry-defining foundations.
tensorflow/tensorflow
Added regression test harness for tf.math.log_sigmoid float64 second derivative calculations.
django/django
Fixed #36523: Contributed django.utils.module_loading.qualname helper to the core Django utility engine.
celery/celery
Fixed database result backend stamping metadata, task children propagation (#8336), and multi-server URL sanitization in bug reports (#8014).
huggingface/transformers
SigLIP2 Flash Attention implementation examples and VibeVoice speech synthesis documentation pipelines.
microsoft/PyRIT
Engineered Garak divergence scenario for automated AI safety testing and red-teaming orchestrator evaluation suites.
langchain-ai/open-swe
Prompt-section configuration mapping for autonomous software engineering agent benchmarks.
Technical Expertise & Stack
Core languages, backend frameworks, and AI systems competencies.
Languages
- • Python 3.12 (Advanced)
- • C / C++
- • SQL (PostgreSQL, MySQL)
- • TypeScript / JavaScript
- • Bash / Shell
Backend & Systems
- • FastAPI & Django REST
- • Celery & Redis Queues
- • Keyset Cursor Pagination
- • Asyncio & Concurrency
- • WebSockets & Event Streams
AI & Machine Learning
- • Google TensorFlow
- • Hugging Face Transformers
- • LangChain & Multi-Agent Workflows
- • LLM Evaluation & Benchmarks
- • RAG Architecture & Vector DBs
DevOps & Infrastructure
- • Docker & Docker Compose
- • GitHub Actions CI/CD
- • Reverse Proxies & Gateways
- • Architecture Decision Records (ADRs)
- • Pytest Automation