Jineshwar Nariani
AI Engineer · Researcher · Builder · Founder
Research in multi-agent reinforcement learning, LLM systems and model evaluation — an honors thesis, a co-authored paper and a paper proposal accepted at IEEE IDS 2025.
Selected Experience
- PRISM — NSF Research2025–2026 · Multi-Agent LLM Systems for Four Research Labs · Retrieval-Grounded Answers, Benchmarked Retrieval
- Manning CICSUndergraduate Researcher · Financial RL + LLMs for Portfolio Optimisation · IEEE IDS 2025 Proposal · AIRL
- UMass Math & Stats REUSummer 2025 · Tariff-Policy Forecasting · Co-Authored Paper · +27% Over Deep-Learning Baselines
Selected Projects
- KausalCausality Tests for World-Model Data · Permutation Tests + FDR · 100% Desync Detection vs 0–30%Also in Software Engineering
- Model’s-Eye TranscodingVLM-Aware Compression · 2,160 VLM Answers · Bootstrap and McNemar Tests Against a Noise FloorAlso in Software Engineering
- LLM 10-K AnalysisLLM Analysis of SEC 10-K Filings, 1995–2023 · Risk, Management Discussion and Sentiment TrendsAlso in Software Engineering
Research
- Honors ThesisStrategic Deception in Financial Markets · Bayesian Detector Beat Deep RL (99.5%) Until the Investor Adapted
- Financial RLPaper Proposal Accepted at IEEE IDS 2025 · MARL, AIRL, Portfolio Optimization
- Tariff-Policy ForecastingCo-Authored Paper · VMD-LSTM, Granger Causality, Time Series, Forecasting
- Kan Extensions for Causal TransferCategory Theory, Causal Inference, LLMs, HPC
- Program Synthesis vs. LLMsProgram Synthesis, Rosette, LLM Evaluation
Education & Honours
- UMass AmherstB.S. Computer Science & Mathematics (Statistics) · GPA 3.86 · 2026
- Dean’s ListAll Eight Semesters · Commonwealth Honors College
- ScholarshipsChancellor’s Award · Dean’s Merit Scholarship · Dean’s International Scholarship
- Student SpeakerManning CICS 2026 Graduation Ceremony
- ASA DataFest 2026Best in Group · 7.7M Patient Encounters · Disproved Our Own Hypothesis, Found a 3.3× ED EffectAlso in GTM / Forward-Deployed