About
I'm an Applied ML Engineer at fal.ai. Previously Lead ML Engineer at PictureStudio, where I built end-to-end multimodal creation platforms. I hold an MS in Computer Science from Rutgers University and a B.Tech from Guru Gobind Singh Indraprastha University.
My work spans the full stack of ML — from training LoRA/ControlNet pipelines that scale across GPU clusters, to building self-serve fine-tuning UIs, to driving R&D on quantization for on-device generation.
Experience
Applied ML Engineer
fal.ai · Remote
- Working on diffusion and autoregressive multimodal models.
Lead ML Engineer
PictureStudio · Remote
- Collaborated on an end-to-end multimodal creation platform — React/Next.js + GraphQL frontend over Python micro-services — that lets artists see, steer, and iterate on diffusion & autoregressive models; now generates 1000+ images per day.
- Designed and productionized a LoRA/ControlNet training + inference pipeline (PyTorch, K8s) that scales from a lone GPU to multi-node clusters.
- Extended open-source frameworks (ComfyUI, A1111) with custom nodes for depth, style, and 3-D camera paths.
- Launched self-serve fine-tuning UI — drag-and-drop asset upload, handling 1–10,000+ images/job while maintaining high subject/style fidelity via VLM auto-captioning and image pre-processing.
- Built a 100k-prompt human + synthetic eval harness; blocks quality regressions pre-deploy and feeds actionable metrics back to Research/Tech-Art.
- Drove R&D on quantization and adapter training, unlocking on-device generation and new editing workflows now adopted by 90% of power users.
Full Stack SWE Intern
Thrasio · NYC, USA
- Ported RESTful APIs to GraphQL to improve reliability and scalability. Developed React components and leveraged Terraform and Kubernetes for automated environment provisioning.
- Built a backend upload service with GraphQL served from S3 for resilient file processing. Implemented async multi-table data updates and Elasticsearch search for fast querying.
- Developed a barcode verification service using TypeScript, React, and PostgreSQL with custom GraphQL APIs and production monitoring.
Backend Software Developer
Shiryam Technologies · India
- Architected modular CMS backend in Python/Flask, deployed on AWS. Reduced asset load times 5x to under 200ms via caching and async processing.
- Developed reusable automated data pipelines for an algorithmic trading platform. Designed resilient DB ingestion mechanisms and failover handling.
- Integrated payment gateways and built out auth flows and hardened KYC processes for security.
ML Intern
BlueStacks · India
- Designed and implemented smart controls recognition system using PyTorch, integrated natively into the BlueStacks application stack. Built scalable ML infrastructure and CI/CD pipelines.
- Achieved 98.78% accuracy and 60ms average latency in real-world tests with a model size of just 10 MB.
- Built an in-game OCR system using DL-based RCNNs. Achieved 77% image-to-text accuracy across English, French, German, Korean, Mandarin, and Japanese.
Skills
Languages
PythonC/C++C#TypeScriptJavaScript
Frameworks
PyTorchSvelteReactNode.jsFlaskDjangoFastAPIPostgreSQL
Developer Tools
GitTerraformDockerKubernetesCircleCIOpenAI APIsGCPAWSAzureLinux
Education
Rutgers University
MS Computer Science
Guru Gobind Singh Indraprastha University
B.Tech Computer Science