Technical leader building ML/AI systems — and shipping product‑quality software.
I’m Shivam Kumar. I’ve led cross‑functional teams delivering RLHF/SFT datasets for LLM training, built evaluation tooling for multi‑modal models, and shipped ML systems that generated real revenue.
About
Technical leader building ML/AI systems and scaling teams. I’ve led cross‑functional teams of 300+ engineers, generated millions in revenue through ML initiatives, and contributed to an international patent. I’m hands‑on across LLMs, computer vision, and full‑stack development.
Focus areas
LLM training data (RLHF/SFT), evaluation tooling for multi‑modal models, automation + QA pipelines, and end‑to‑end systems that scale.
Education
B.Tech (Computer Science), IIT Goa (2020) · CPI: 9.22/10.0
Experience
Roles and impact highlights.
Delivery Manager / Technical Lead · Turing (Remote, Contract)
Led and scaled cross‑functional teams of 300+ engineers delivering RLHF and SFT datasets for LLM training with top AI firms and FAANG companies. Built scalable operations from POC to 450+ people across parallel projects.
Designed performance tracking pipelines and automated QA systems with CI/CD to ensure high‑quality delivery and support expanded multi‑million dollar engagements.
Built VLMBench evaluation tooling for multi‑modal language models (23 models, 705 samples) using GCP and LLM APIs (OpenAI/Gemini) with a methodology for comparing textual predictions to ground truth.
Designed an Agentic Reviewer system for automated code review — reducing manual review workload while improving quality.
Web3 Developer / Founder · Multiple projects (Solana, NEAR)
Built NFT staking, raffle, and auction applications on Solana and NEAR using Rust and TypeScript. Engineered a pioneering Discord bot for NFT holders (first on NEAR) and Solana wallet auth inside Unity games.
Created Mars Minter, an open‑source minting solution on NEAR with CLI tools and frontend templates.
Software Engineer I · EagleView (Bengaluru, India)
Developed an ML model for Digital Surface Model (DSM) generation from images — resulting in a published patent and enabling operations in regions without reliable elevation data.
Automated ~80% of solar report alignment using computer vision (OpenCV, PyTorch) and built a 2D preview algorithm reducing preview time from ~10 minutes to ~20 seconds (30×).
Architected scalable image data collection + preprocessing pipelines in Python on GCP for large‑scale model training.
AI Engineer + Full Stack Developer · Learngram (Bengaluru, India)
Built an AI voice cloning model using NLP to enable personalized learning. Shipped an ed‑tech platform with a React Native mobile app and Python backend, including video calls, assignments, and calendar integration.
Research Intern · NTU (Singapore)
Built an ML-based engine for binary vulnerability detection using hierarchical filtering and locality sensitive hashing, achieving >90% detection accuracy.
Patents
Selected patent work.
DSM Reconstruction using Artificial Intelligence
Co‑invented an ML system generating 3D Digital Surface Models (DSMs) from images using computer vision, reducing reliance on LiDAR and enabling operations where elevation data was unavailable.
Scholastic highlights
Technical Excellence Award (IIT Goa, 2018–2019) · JEE Advanced AIR 3120 (2016)
Projects
A few public artifacts and builds.
Mars Minter
CLI tools + frontend templates to accelerate NFT minting launches on NEAR.
VLMBench (evaluation tooling)
Built evaluation tooling to compare multi‑modal model predictions against ground truth using cloud + LLM APIs.
Skills
Core stack and tools.
Programming
Frameworks & libraries
ML/AI
Cloud, DevOps, Web3
Leadership & responsibility
Teaching Assistant (Foundations of ML, Computer Programming — IIT Goa, 2017–2020) · Internship Coordinator (Placement cell, IIT Goa, Oct 2017 – Sep 2018)
Methodologies
Agile/Scrum · performance tracking · QA systems · cross‑functional collaboration · system architecture