ayush / stochi

I like World Models, RL & AI for Science.

Prev: RL Resident @PrimeIntellect

See you at NeurIPS 2026 in Sydney.

Latest work

  • You Don’t Need Astra to Edit a Model’s Weights

    —

    A GPT-OSS follow-up to Weight-Space Programming: direct edits to a small LunarLander policy, with held-out evaluation.

  • Intervention Search & Response Prediction

    —

    An agent found an intervention that worked. Could it predict what would happen with one it hadn’t tried?

    AgenticLS · NeurIPS 2026 workshop

    Accepted spotlight
  • Budgeted Predictive Knowledge

    —

    When the world changes and an agent can only recheck a few predictions, which should it check first?

    NEmo · NeurIPS 2026 workshop

    Accepted poster

Work & affiliations

Prime Intellect · RL ResidencyMar–Jun 2026

San Francisco, USA

  • Built Statecraft to evaluate tool use in mutable worlds; ran Sparse Signal Loop to compare feedback and memory choices in agent refinement.
  • Developed RL environments and long-context reasoning evaluations, including LongCoT and RLM variants.
QX Labs · Founding AI EngineerJun–Oct 2025

London, UK

  • Built a knowledge graph and agent orchestration for private-equity research, ingesting approximately 3 million company records.
  • Reduced recommendation and query latency from 42 seconds to 10–12 seconds.
Covenants PharmaChem · Tech LeadOct 2024–Jul 2025
  • Built an AI platform for procurement automation.
Unsiloed AI (YC F25) · Founding AI EngineerJan–Jun 2025

San Francisco, CA

  • Built streaming document-extraction APIs and evaluated field-extraction models on 10,000+ financial reports and SEC filings.
  • Fine-tuned YOLOv11 for document layouts and optimized inference with TensorRT, ONNX, and OpenVINO.
IIT Bombay · Research InternAug 2024–Jan 2025
  • Implemented spiking neural networks for German speech classification in the Spin-Neuro-Quantum-AI Group.
  • Built real-time fault prediction for electromagnetic machines using signal and magnetic-flux analysis.
Earlier work 3 roles
Bachelor’s — Computer Engineering · Teaching Assistant, Advanced Algorithms and ComplexitySept–Dec 2024
  • Led problem-solving sessions and designed research-based problem sets for 16 master’s students.
IEEE CS Student Chapter · CoordinatorNov 2022–Dec 2023
  • Coordinated chapter events, conferences, hackathons, and competitions.
Predoole Analytics · Developer Analyst InternJun–Aug 2023
  • Built a personalized email-generation API integrating Apollo.io data.