I am currently a Master of Finance student at MIT. I previously completed my Bachelor of Science in Computer Science at Columbia University, where I focused on machine learning, algorithms, and applied data analysis.
A collection of statistical and machine-learning models for analyzing and forecasting financial time-series data.
Tools & Technologies:
Python, NumPy, pandas, SciPy, statsmodels, arch, scikit-learn, Jupyter notebooks, matplotlib/seaborn
Concepts & Methods Used:
- Time-series modeling (AR, MA, ARMA, ARIMA)
- Stationarity testing (ADF/KPSS), autocorrelation analysis
- Likelihood-based estimation (MLE) for GARCH models
- State-space modeling and Kalman filtering
- Vector Autoregression (VAR), impulse responses, variance decomposition
- Rolling-window evaluation and out-of-sample forecasting
- Model selection (AIC/BIC), residual diagnostics, forecast error analysis
Developed an ensemble of transformer-based and foundation models to predict 3D RNA geometry.
Tools & Technologies:
PyTorch, pandas, NumPy, scikit-learn, GPU-accelerated training
Concepts & Methods Used:
- Transformer architectures and attention mechanisms
- Sequence modeling for high-dimensional biological data
- Data preprocessing, batching, and normalization
- Model training pipelines using PyTorch’s autograd and DataLoader utilities
- Cross-validation and fold management
- Ensemble modeling for variance reduction and stability
Ongoing work to strengthen algorithmic problem-solving ability.
Tools & Technologies:
C++, Python, STL (Standard Template Library), custom algorithm templates
Concepts & Methods Used:
- Algorithm design (dynamic programming, divide-and-conquer, greedy methods)
- Graph algorithms (BFS/DFS, Dijkstra’s, tree algorithms, topological sorting)
- Classical data structures (segment trees, Fenwick trees/BIT, hash maps, priority queues)
- Mathematical algorithms (modular arithmetic, combinatorics, binary lifting)
- Complexity analysis (time and space, amortized cost)
- Writing optimized solutions for performance-critical environments
A system designed to support real-time conversational interaction.
Tools & Technologies:
React, Node.js, Django, Python, multithreading, REST APIs, WebSockets, speech-to-text and text-to-speech libraries
Concepts & Methods Used:
- Concurrent programming and thread-based parallelism
- Asynchronous request handling and real-time data streaming
- Backend systems design for low-latency applications
- API design, session/state management, structured data flows
- Integration of external ML components (LLMs, voice models)
- High-throughput inference and performance tuning
Email: nikhibel@mit.edu
LinkedIn: https://linkedin.com/in/nikhilesh-belulkar