Subhaditya Mukherjee SubhadityaMukherjee

sequenceDiagram

participant User as Intent Recognition

participant Controller

participant AutoML as AutoML Engine

participant Tabular as Tabular Engine

flowchart LR
  %% Intent and Control
  A[Intent Recognition] -->|Extracts task type & user input| B[Controller]
  B -->|Requests requirements| C[AutoML Engine]
  C -->|Returns requirements JSON| B
  B -->|Sends config, data, task info| C

  %% Core AutoML Tasks
  subgraph Core_AutoML_Tasks["Core AutoML Tasks"]

Links

Presenter Info

Subhaditya Mukhejee : [email protected]

Freelance Projects/Blogs I have worked on

I am Subhaditya, it's nice to meet you!
Below is a list of projects I have worked on for various clients over the years.
I wanted to keep a list of them for both myself and future employers. 
That being said, I am almost always open for more freelancw work, so hit me up if you want to discuss some projects :)

While most of the code is closed source due to client requirements, I am still open to discussing them to a certain extent
My Github is https://github.com/SubhadityaMukherjee

	"""
	Run result aggregator
	- Read all tensorbord logs and save as a pandas dataframe for analysis
	"""

	import os
	import pandas as pd
	from tensorboard.backend.event_processing.event_accumulator import EventAccumulator
	from tqdm import tqdm
	from pathlib import Path

	import pickle
	with open("pickled_df.pkl", "wb+") as f:
	pickle.dump(combined_df, f)

	with open("pickled_df.pkl", "rb+") as f:
	combined_df = pickle.load(f)

	def process_event_acc(event_acc, save_ims=False) -> dict:
	"""Process the EventAccumulator and return a dictionary of tag values"""
	all_tags = event_acc.Tags() # Get all tags
	temp_dict = {} # Store all values here
	for tag in all_tags.keys(): # Loop over all tags
	if tag == "scalars":
	# Process scalars
	for subtag in all_tags[tag]:
	try:
	# Try to get the last value

	def process_runs(main_path, save_ims=False) -> pd.DataFrame:
	"""Process all runs and return a dataframe of all results"""
	all_files = get_event_files(main_path=main_path, save_ims=False)
	all_dict = {}
	for files in tqdm(all_files, total=len(all_files)):
	try:
	# Process each file using the EventAccumulator and save to a dictionary
	event_acc = EventAccumulator(files)
	event_acc.Reload()
	temp_dict = process_event_acc(event_acc, save_ims=save_ims)

	filtered_df = combined_df[(~pd.isnull(combined_df['converted_proxy'])) & (~pd.isnull(combined_df['original_images']))]
	filtered_df.iloc[0].original_images

	combined_df = process_runs(main_path=main_path)
	combined_df = combined_df[(~pd.isnull(combined_df['experiment_name'])) & (~pd.isnull(combined_df['Loss/Val']))]
	combined_df.head()