Mohammad Ausaf Logo Image
Mohammad Ausaf

mohammad ausaf

GenAI · Backend · Infra
@ Galleri5 (Collective Artists Network)

get to know me

I'm a software engineer based in Bengaluru, working at Galleri5 (part of Collective Artists Network) where I build generative AI systems, backend infrastructure, and the tooling around them. I joined when the tech team was around 10 people, and as one of four engineers on the AI stack, I've had the chance to work across the full depth — from ML research and model training to distributed systems and production operations.

Most of my day-to-day revolves around agentic AI — autonomous multi-step workflows that decompose creative prompts into execution DAGs, select and orchestrate models across providers, and stream results in real time. On the infrastructure side, I've built the distributed job dispatch system, GPU fleet management, and the concurrency and crash recovery layers underneath it all. I also handle deployments across GCP, Azure, and AWS, and lead on-call rotation.

The pipelines I've built have powered actual shipped content — Mahabharat - Ek Dharmayudh streaming on JioHotstar, and the AI-generated teaser for Chiranjeevi Hanuman: The Eternal (8M+ on YouTube / 12M+ on Instagram, premiered on IMAX). Currently co-developing an AI asset platform with Microsoft.

Before generative AI, I worked on classical ML — NLP, clustering, vector search — powering trend-based product discovery for brands like Myntra and Ajio. On the R&D side, I build ComfyUI workflows, train LoRAs for character consistency and style transfer, and develop new generation pipelines from research through to production.

Email Me

Tech Stack

Python
FastAPI
PyTorch
MongoDB
Redis
Docker
GCP
Azure
AWS
Firebase
Pinecone
ComfyUI
CLIP
LangChain
Google Gemini
OpenAI
Replicate
FAL AI
Agentic AI
Generative AI
Distributed Systems

experience

software engineer (ai)

Galleri5 - Bengaluru

Dec 2023 - Present

  • Agentic AI Platform: Led development of AI Studio - multi-tenant content creation platform with RBAC and enterprise permissions. Built Workflow Planner using Gemini for natural language interpretation and multi-step execution plans with dependency resolution
  • Microsoft Collaboration: Co-developing AI asset platform with Microsoft. Designed multi-layer architecture handling versioned storage, semantic retrieval, automated quality evaluation, and preference-aware ranking. Implemented real-time review workflows with human feedback integration for creative iteration
  • Distributed Infrastructure: Built Inference Gateway orchestrating 78 models across 7 providers with WRR-based fair scheduling and 4-level concurrency control. Designed GPU Dispatcher managing 8 H100 GPU server fleet with lease-based locking, priority queuing (Redis sorted sets), and crash recovery via WAL
  • ML Pipelines: Built end-to-end ML pipelines serving major e-commerce clients (Myntra, Ajio, H&M). Developed CLIP-based vectorization for semantic product search across millions of SKUs, text classification with sentence transformers (BGE, mxbai-embed), and real-time vector upsertion with Pinecone
  • R&D & Prototyping: Researched and prototyped SOTA image/video generation techniques. Built custom ComfyUI workflows for client-specific image generation. Developed image generation evaluation loops using Gemini Vision for automated quality assessment
  • Team Leadership: Technical lead for 3-person junior engineering team. Established code review processes, JIRA workflows, and onboarding practices

Tech Stack: Python, FastAPI, MongoDB, Redis, AWS, GCP, Azure, Firebase, PyTorch, CLIP, Pinecone, ComfyUI, Docker, LangChain, OpenAI, Gemini, Replicate, FAL AI

bachelor of technology - computer science

KIET Group of Institution, Ghaziabad

2024

Relevant Coursework:

  • Probability and Statistics, Calculus, Operating Systems
  • Data Structures and Algorithms, Machine Learning
  • Databases

work

Tinify Platform Screenshot

ai studio - agentic ai content platform

Enterprise agentic AI platform where autonomous agents orchestrate complex multi-modal content workflows. Features intelligent workflow planning with Gemini, real-time collaboration systems, and dynamic task execution across 78 models and 7 providers. Technologies: FastAPI, MongoDB, LangChain, Redis, Multi-Agent Systems, Computer Vision.

Read More
Searchy Screenshot

searchy - on-device ai photo manager

Native macOS photo management app with AI-powered features, entirely on-device with zero cloud dependency. Features CLIP-based semantic search, duplicate detection with visual similarity scoring, face recognition and grouping, smart filtering, and auto-indexing with directory watchers. Modular architecture with swappable ML backends. Technologies: Swift, SwiftUI, CLIP, Face Recognition, Metal (GPU).

View Project
Boring Notch Lyrics

boring.notch lyrics - macOS notch music widget

Fork of Boring Notch adding real-time synchronized lyrics display. Features multiple display modes (flowing, alternating, stacked), per-monitor configuration, and timing offset controls. Supports Spotify, Apple Music, and YouTube Music. Technologies: Swift, SwiftUI, macOS APIs, Music Integration.

Read More
Deep Learning Plant Detection

deep learning based medicinal plants detection

This project aimed to identify medicinal herbs using a machine learning model(s). The model used ResNet, a type of CNN, with a validation accuracy of 96%.

Read More
Software Screenshot

cv-repcounter-timer

This project involved using Media Pipe's pose detection system to count the reps done by a subject. The model calculates the angle between body parts and registers a recount based on a threshold with special emphasis on filtering fluctuation of feed in real time.

Read More