Sokhal is Moshi’s founder and lead machine-learning engineer. The company builds 24/7 agentic quantitative analysts for financial markets and serves institutional, fund, and corporate pilot customers.
Sokhal founded JewelVision and led its AI engineering in 2024. The project developed multimodal-AI tools for 3D asset generation and augmented-reality virtual try-on in e-commerce.
Sokhal participated as an AI/ML engineer in this Rice D2K capstone with NASA Johnson Space Center. The team built an efficient lunar-crater detection pipeline combining object detection and ellipse fitting to support autonomous spacecraft navigation on constrained hardware.
An ICLR 2026 publication co-authored by Anekha Sokhal. The project benchmarks 3D geometric reasoning in vision foundation models using simulated and real polyhedra and reports that the evaluated models struggle with basic geometric reasoning.
Anekha Sokhal is the founder and CEO of Moshi, a Houston-based company building real-time, cross-asset intelligence for institutional investors. Her approach to financial AI draws directly on an unusual career sequence: seven years in commodities and energy derivatives, a move into data science and computer vision, and work spanning e-commerce, lunar navigation, and 3D geometric reasoning.
From energy trading to data science
Sokhal earned a degree in Mathematics, Operational Research, Statistics and Economics from the University of Warwick in 2014. She joined J.P. Morgan that year, initially rotating through emerging-markets roles before becoming a commodity analyst and, in 2017, an energy-derivatives trader. Her public résumé describes work building forecasting models for European gas markets, structuring and executing energy derivatives, and automating real-time risk dashboards.
During the pandemic, Sokhal began applying data analysis to her family’s jewellery business. The contrast with trading changed the direction of her career. Energy desks routinely used data to manage risk and guide decisions; in retail, she saw businesses operating without comparable analytical systems. She subsequently led strategy and data analytics at ESKA International and completed data-science training at Le Wagon.
A Fulbright Scholarship brought her to Rice University, where she completed a professional master’s degree in data science with a specialization in machine learning and computer vision. Rice’s Fulbright profile traces her transition from quantitative finance into data-driven retail strategy, while her later technical work shows how she extended those analytical habits into deployable AI systems.
Computer vision for products, navigation, and geometry
In 2024, Sokhal founded JewelVision, which developed 3D asset-generation and augmented-reality virtual try-on tools for e-commerce. The project joined her experience in jewellery with multimodal AI, addressing the problem of reconstructing and presenting products whose geometry is difficult to convey through ordinary photographs.
At Rice, she also worked as an AI and machine-learning engineer on a capstone with NASA Johnson Space Center. The team’s open-source LunarLens crater-detection system combines object detection with ellipse fitting to identify crater rims under difficult lighting and perspective conditions. Because the system was intended to support autonomous navigation beyond GPS coverage, the project treated efficient operation on low-power spaceflight hardware as part of the research problem.
Sokhal later co-authored GIQ, an ICLR 2026 paper that benchmarks 3D geometric reasoning in vision foundation models using simulated and real polyhedra. The work examines whether models can reason about spatial structure rather than merely recognize the objects shown to them.
Moshi’s division of labor between models
Sokhal returned to finance when she founded Moshi in 2025. Mercury Fund describes Moshi as a real-time agentic AI platform that gives institutional investors cross-asset intelligence. Sokhal describes its product as a set of 24/7 agentic quant analysts designed to monitor markets and identify developments relevant to a user’s positions.
Her product philosophy begins with what she calls prompting fatigue: the cycle of writing instructions, reviewing a generic answer, identifying what it missed, and trying again. In her explanation of Moshi’s architecture, she argues that requiring a trader to supervise this process does not remove work from the desk; it transfers the work to the person whose judgment is most expensive.
Moshi therefore assigns different jobs to different kinds of models. Sokhal says its relevance analysis relies on deterministic models built around statistics, market structure, and domain knowledge. A language model operates at the presentation layer, turning the resulting analysis into readable output. The distinction captures the practical lesson running through her career: use quantitative systems to perform the analysis they can structure reliably, use language models where language is the task, and keep the investor focused on the decision rather than the machinery producing it.
Anekha Sokhal explains why financial AI should absorb the market-data grind, expose sources and assumptions, and leave consequential trading judgment with the investor.
Investor edge combines earlier awareness, detection of second- and third-order effects, and enough conviction to size a trade.
A production financial system must solve security, storage, reliability, current-data, cost, and accountability problems that a weekend dashboard can ignore.
The useful human-machine boundary assigns collection, cleaning, monitoring, and cross-checking to machines while leaving consequential judgment with the investor.
The hard open problem is not gathering more data; it is determining how much tacit investor context can be represented without erasing human responsibility or apprenticeship.
