Headline

Founding Engineer | ML Researcher at Yale | Open to AI/ML and founding engineer roles

About

I build AI products and ML systems—from model internals to production infrastructure. As Founding Engineer at Uniiq.ai, I own an AWS-deployed admissions platform serving 250+ users. I rebuilt conversational onboarding with structured LLM outputs, retries, rate limits, persisted state, and role-aware API protection; strengthened data/admin workflows; and improved Lighthouse performance from 73→93, LCP from 3.3s→0.7s, and INP to 130ms. Previously, at Yale University, I developed privacy-preserving CNN components from first principles in Python/NumPy for Multi-Party Computation, achieving 88.08% accuracy on ChestMNIST. I also publish tested, from-scratch ML systems: a Whisper-style ASR stack, a PEFT library, a Rust ONNX inference server, and an HNSW vector index. Earlier, I automated 15+ quantitative-research workflows at WorldQuant and ranked Top 1% globally in the Global Alphathon. Focus: Applied AI, ML systems, LLM inference, retrieval, secure AI, PyTorch, FastAPI, TypeScript, and Rust.

Location

Mumbai

AI stack

claude code

claude

codex

opencode

Work experience

Founding Engineer at Uniiq, May 2026 - Present · 4 mos, Mumbai, AI-powered university admissions platform helping students discover universities, build application profiles, and manage admissions decisions. - Led engineering for an AWS-deployed AI admissions platform, owning architecture, AI integrations, frontend, backend, infrastructure, and deployments. - Rebuilt onboarding as a 3-phase AI conversation with adaptive follow-ups, transcript persistence, structured validation, rate limiting, and recoverable Gemini failures. - Implemented LLM security controls, role-aware API protection, data-integrity safeguards, and admin workflows for institutions and opportunities. - Improved performance score from 73→93, reduced LCP from 3.3s→0.7s, and achieved 130ms INP through code splitting, bundle optimization, compression, and caching.

Dev Ambassador at MiniMax, Nov 2025 - Feb 2026 · 4 mos, - Built technical proofs-of-concept and API integrations demonstrating generative AI platform capabilities. - Created developer documentation and tutorials to reduce integration friction and improve developer onboarding. - Served as a technical liaison between external developers and the product team, converting community feedback into actionable developer-experience improvements.

Machine Learning Researcher at Yale University, Oct 2025 - Mar 2026 · 6 mos, Remote, - Developed privacy-preserving deep-learning components from first principles in Python/NumPy, without external autograd, for training CNNs under Multi-Party Computation. - Designed a custom CNN that achieved 88.08% test accuracy on ChestMNIST while operating over cryptographically protected data. - Authored Sequre contributions covering Conv2D, CNN layers, encrypted pooling tests, and a ChestMNIST MPC training pipeline.

Machine Learning Researcher at AI Pioneers (aipioneers.me), Oct 2025 - Mar 2026 · 6 mos, Remote, - Developed privacy-preserving deep-learning components from first principles in Python/NumPy, without external autograd, for training CNNs under Multi-Party Computation. - Designed a custom CNN that achieved 88.08% test accuracy on ChestMNIST while operating over cryptographically protected data. - Authored Sequre contributions covering Conv2D, CNN layers, encrypted pooling tests, and a ChestMNIST MPC training pipeline.

Quant Research Consultant at WorldQuant, Nov 2022 - Jul 2025 · 2 yrs 9 mos, - Automated 15+ quantitative signal-generation and backtesting workflows using Python across multiple global markets. - Developed strategies with a documented average simulated Sharpe ratio of 1.8 and ranked Top 1% globally in the 2022 Global Alphathon.

Education

Bachelor of Engineering - BE, Computer Science, University of Mumbai

Global Summer School, Quantum Machine Learning, Quantum Simulation, Qiskit (by IBM Quantum)

Public links

github: https://github.com/sdivyanshu90