Experience — Jineshwar Nariani
PRISM — NSF Research
AI Software Engineer · Jan 2025 – Jun 2026 · Amherst, MA
Worked in a near forward-deployed role across a gene lab, a biotech lab, a photonics lab and the campus library’s resource office, building one multi-agent research framework re-pointed at each. Role-specialised agents (retrieval, analysis, synthesis) ran inside LangGraph graphs for branching and retries; each lab’s data came in through its own MCP server, so onboarding a new lab meant a new tool server, not new orchestration. Answers were retrieval-grounded, and retrieval strategies shipped only after winning feature-flagged benchmarks. The specifics differed lab to lab, but the underlying mechanism was common — each new lab went faster than the last.
Agents, LangGraph, n8n, MCP, RAG, Evals, Docker
Manning College of Information & Computer Sciences
Undergraduate Researcher · Jan 2025 – May 2026 · Amherst, MA
Designed a financial reinforcement-learning architecture combining multi-agent RL with large language models for personalised portfolio optimisation — paper proposal accepted at IEEE IDS 2025. Implemented adversarial inverse reinforcement learning (AIRL) for market-making simulations.
Multi-agent RL, AIRL, PyTorch, LLMs, Finance
UMass Amherst Mathematics & Statistics
Summer REU · Jun – Aug 2025 · Cambridge, MA
Analysed 22 years of U.S. tariff policy against trade flows and macro indicators (correlation, mutual information, Granger causality, random forests). Built a hybrid VMD-LSTM forecaster that beat LSTM, BiLSTM and GRU baselines by 27%; co-authored the resulting paper.
Time series, Statistics, Economics, Machine Learning, Data Science
Sky Governance
Backend Engineer · Jun – Aug 2024 · Berkeley, CA
Designed a microservices architecture integrating SQL databases across AWS, Azure and GCP, with retrieval-augmented fine-tuning (RAFT) for anomaly detection and compliance monitoring — real-time alerts and automated remediation of governance breaches.
Microservices, Databases, AWS, Azure, GCP, RAFT
Accurate Industrial Controls
Machine Learning Intern · Jun – Aug 2023 · Pune, India
Trained decision-tree and CNN models on real-time sensor data from industrial robots that manufacture electrical transformers, and modelled the robots’ physics in MATLAB and Python — improving robotic-arm efficiency by 20%.
TensorFlow, Keras, MATLAB, SciPy, NumPy, Machine Learning, Robotics