ML Intern @ Eachlabs.ai (May-Sep 2025)

This was my first industry experience. I was a complete newbie with some experience in LLMs, but this role was pretty exciting as i got to work with generative models. my CTO and manager were really nice and taught me a lot of new things (soft skills too)

I started by working on image face swap and took complete ownership of that project. Our face swap service was the fastest and cheapest in the market. Later, I also shipped video face swap. None of our competitors (Fal, Replicate, etc.) had it at that time.

Learnings: ONNX and TensorRT to accelerate inference.

After that, I slowly dived into model optimization. I read a hella lot of tutorials and GitHub discussions, learned a lot of new things to reduce latency, and tried out different techniques. The ones that gave the most returns were quantization, caching, and frame interpolation (for video models).

Biggest takeaway: For startups, speed matters a lot.

---