johnzakkam

Location: Hyderabad, India Remote: Yes

Willing to relocate: Yes

Technologies: Python, PyTorch, ONNX, ONNX Runtime, quantization, QDQ, INT8/INT4, model compression, calibration, accuracy recovery, computer vision, LLM fine-tuning, LoRA/QLoRA, evaluation pipelines, Docker, Linux

Résumé/CV: Available on request

Email: johnzakkam2592@gmail.com

AI / ML engineer focused on model optimization, quantization, and deployment.

I help teams take models from “works in research” to “runs efficiently in production.” My work includes PyTorch to ONNX export/debugging, INT8/INT4 quantization, calibration, accuracy recovery after compression, latency/memory benchmarking, and fine-tuning/evaluation workflows.

Best fit: startups or teams with ML/AI models that are too slow, too expensive to serve, hard to deploy, or losing accuracy after quantization/compression.

Interested in freelance, consulting, part-time, or full-time opportunities around model optimization, edge AI, inference efficiency, and applied ML systems.

Location: Hyderabad, India

Remote: Yes

Willing to relocate: Yes

Technologies: Python, PyTorch, ONNX, ONNX Runtime, quantization, QDQ, INT8/INT4, model compression, calibration, accuracy recovery, computer vision, LLM fine-tuning, LoRA/QLoRA, evaluation pipelines, Docker, Linux

Résumé/CV: Available on request

Email: johnzakkam2592@gmail.com

AI / ML engineer focused on model optimization, quantization, and deployment.

I help teams take models from “works in research” to “runs efficiently in production.” My work includes PyTorch to ONNX export/debugging, INT8/INT4 quantization, calibration, accuracy recovery after compression, latency/memory benchmarking, and fine-tuning/evaluation workflows.

Best fit: startups or teams with ML/AI models that are too slow, too expensive to serve, hard to deploy, or losing accuracy after quantization/compression.

Interested in freelance, consulting, part-time, or full-time opportunities around model optimization, edge AI, inference efficiency, and applied ML systems.

This is incredibly valuable advice, thank you! You're absolutely right that nothing replaces real investor feedback - the skin in the game factor is huge. Your process of working up from least preferred to most preferred investors is brilliant and something I wish I'd known earlier.

I'm thinking the tool might be more useful in the prep phase, before you even reach out to that first investor on your list. The goal wouldn't be to replace your process, but maybe help founders feel more confident before that first real conversation? Especially founders who don't have experienced networks to practice with initially.

Does that positioning make more sense, or do you think even that prep stage isn't worth the effort vs just jumping straight into real conversations?

Interesting! That's exactly the kind of insight I was hoping for.

Just checked out your profile, you probably had access to other experienced founders who'd been through similar fundraising experiences. Were they more difficult because of that direct experience, or just naturally more critical in their review style?

I'm thinking most early founders don't have that kind of network though, maybe that's why they're getting much softer feedback.

Did you find the VC questions predictable after getting grilled by fellow founders, or were there still surprises?