MP/now
CURRENT STATE · UPDATED OCT 2026

Now

What I'm working on as of October 26, 2026. I keep this short and update it when the priorities change. Format borrowed from Derek Sivers.

PAGE · nowUPDATED · oct 26, 2026

Current

  • First year at UofT. Studying in Arts & Science at St. George, intending to pursue the Applied Mathematics Specialist.
  • AI research. Moving my focus toward machine learning, model evaluation, and efficient inference. I'm looking for collaborative projects where I can own experiments and learn from experienced researchers.
  • Open source. Seeking larger AI/ML contributions with responsibility for methods, implementation, benchmarks, and experimental validation.
  • Existing work. Continuing to maintain my quantitative-research and broker tools while keeping new commitments manageable alongside coursework.

Recently finished

  • JAX #41960 merged October 26: the binom, nbinom, geom and bernoulli log-pmfs in jax.scipy.stats now return -inf at a non-integer k instead of a finite value.
  • MLX #4748 merged October 25, my first MLX pull request: vmap of tri_inv now respects the triangular flags instead of silently returning the full inverse.
  • NautilusTrader #5432 merged October 25: statistics now reach Python in name order, so tearsheet rows stay in place between runs of the same backtest.
  • ib_async: my fix to the Ticker.news example in the v3 migration guide landed in the 3.0 line as my commit on October 23.
  • NautilusTrader #5407 merged October 21: the Rust backtest engine now keeps an instrument's matching engine and orders when it is registered again, treating an identical definition as a no-op and rejecting a changed one.
  • QuantLib-SWIG #928 merged September 28: variance swaps and replicating / Monte Carlo pricing engines exposed to Python, with numerical tests.
  • Previously merged forecasting-metric work in Darts and Rust/Python correctness and statistics work in NautilusTrader.

Evidence boundary

The trading projects are personal research and infrastructure. I do not use private backtest returns as a public credential. The inspectable work is the code, tests, documentation, upstream reviews, and explicit limitations.

Next

  • Join an ML project with a concrete research question, useful mentorship, and room for substantial technical ownership.
  • Build an inspectable AI contribution with strong baselines, ablations, and reproducible results.
  • Find paid technical work and internships while protecting time for university.

mahimn · now · last refreshed oct 26, 2026