-
Gesture Controlled Robot
Course: Mobile Computing and IOT
[
Code]
- Developed a palm-mounted inertial measurement unit (IMU) using sensor data
(accelerometer/gyroscope) with AHRS filtering to convert raw readings into stable
roll/pitch/yaw orientation values. Implemented noise reduction through complementary filtering
()
and magnetic interference mitigation using gyro-derived roll compensation.
- Built an ESP32-to-Raspberry Pi communication system using ESP-NOW protocol with 120ms
round-trip latency, translating filtered orientation data into Sphero SDK commands through
coordinate frame transformation ().
-
Ranked Retrieval System
Course: Information Retrieval
[
Code],
[
Report]
- As part of the project, we developed a vector-space information retrieval system using TF-IDF cosine similarity
to rank 6,362 Wikipedia pages, achieving an initial 25% average accuracy.
- We further made two enhancements—Query Relaxation (synonyms/co-hyponyms) and Latent Semantic Indexing—significantly
improved performance, with some queries reaching 100% accuracy through semantic understanding and expanded
term relationships.
- These were some initial attempts to replicate Google Search xD. IR was a good course though.
-
Language Design and Type Expression Computation
Course: Principals of Programming Languages
[
Code]
-
Designed a context-free grammar for a C-inspired custom language with
native support for jagged arrays and array arithmetic, boolean primitives, operator overloading capabilities,
type-safe array operations and multidimensional data structures.
- Built a compiler toolchain including a recursive descent parser generating abstract syntax trees,
type expression tables for semantic analysis, and error detection mechanisms during parse tree
traversal - implementing memory-efficient storage strategies for irregular array structures
- This was an implementation heavy project in C, but helped me develop a fair understanding of
how exactly a user written code block is parsed, and transcribed into structured representations (ASTs)
and machine-executable logic, how typed languages work and tokenization principles.
-
Statistical Analysis and Forecasting of Solar Energy
Course: Applied Statistical Methods
[
Code],
[
Report]
-
Developed a multivariate statistical methodology to analyze solar irradiance patterns across Indian states,
identifying Global Horizontal Irradiance (GHI) as the most critical solar parameter through correlation
analysis (r=0.97 with Clearsky GHI). Implemented distribution fitting using AIC/KS tests,
determining log-gamma distribution (AIC=1513) as optimal for GHI data modeling,
while decomposing time series to reveal significant annual seasonality through additive decomposition
(Yt = St + Tt + Rt).
-
Engineered a Seasonal ARIMA (1,1,1)(1,1,1,52) model achieving state-specific solar forecasts
with 4.77-10.39% MAPE accuracy, outperforming baseline ARIMA by 32% through grid-searched parameters
and seasonal differencing. Validated the system across four states, with Rajasthan showing optimal
performance (RMSE=42.18), providing actionable insights for solar park siting and grid integration strategies
-
The Student Union Application
SUTT Team
[
App]
-
As part of SUTT, we built the SU App which caters to all the student needs in a quick, simple and efficient manner.
With integration of loads of features: real time expense review system, merchandise
signings, QR based delivery, academic resources, all in a single app. The app was used by 4000+ students at BITS.
-
During the initial years of my undergrad, I built couple of android applications - a betting app for our sports fest
(Code),
a food delivery app (Code) and some more.
