Ankit Khedia

Data Analyst

https://www.opentalent.in/ankit-khedia

Key Strengths

Extensive experience in Machine Learning and Deep Learning, particularly in NLP, Generative AI, and predictive modeling, which are highly relevant to Data Analyst roles involving advanced analytics.
Strong background in data processing, feature engineering, and model deployment from roles at Google and Amazon, indicating practical skills in the data lifecycle.
Experience with various data analysis techniques, including statistical modeling (Gaussian Mixture Models, Monte Carlo simulations), time series analysis (Dynamic Time Warping, LSTMs), and recommendation systems (collaborative filtering).
Demonstrated ability to work with large datasets and optimize data pipelines (e.g., reducing SLA from 10 days to minutes at Amazon).
Academic background from top-tier institutions (Georgia Institute of Technology, IIT Kharagpur) with a Master's in Computer Science.

Cultural & Operational Fit

Cultural Fit Analysis

The candidate's project history shows a broad interest in various domains, from finance to image processing and natural language processing, indicating intellectual curiosity and adaptability. The experience at major tech companies (Meta, Google, Amazon, Oracle, Goldman Sachs) suggests an ability to thrive in structured, high-performance environments. The personal projects, especially those involving market simulation and financial manipulation detection, align with roles requiring analytical rigor and ethical considerations. The target role of 'Data Analyst' is well-supported by the candidate's extensive experience in data analysis, machine learning, and predictive modeling, suggesting a strong cultural and technical fit for data-driven organizations.

Soft Skills & Operational Fit

The candidate's project descriptions suggest a strong problem-solving orientation and an ability to work on complex, multi-faceted problems. The experience as a Graduate Teaching Assistant indicates some communication and mentorship skills. However, without direct assessment data on soft skills or operational fit, a comprehensive evaluation is limited. The diverse project portfolio suggests adaptability and a proactive learning attitude.

AI is analyzing your overall score…

Identifying your key strengths…

Evaluating your skill match against the job requirements…

Assessing your cultural and operational fit

About

I’m Ankit Khedia, a Lead AI engineer at Meta working on generative AI tools using diffusion models, vision-language systems, and LLMs. I've previously built ML systems at Google, Amazon, and Goldman Sachs across search, recommendations, and real-time risk. Beyond engineering, I follow early-stage startups on AI product and strategy and invest in deep tech ventures. I'm passionate about the future of AI in creative, healthcare, and enterprise domains—and love solving hard technical and product problems. Let’s connect if you're building something ambitious in AI and looking for thought partnership/advising or consulting opportunities on early-stage tech!

Top Skills

C++C++MatlabJavaProgrammingMicrosoft OfficeMicrosoft ExcelWindowsData StructuresMicrosoft WordPowerPointLinuxAlgorithmsResearchMysqlnetbeansEnglishPhotoshopData MiningMachine LearningNatural Language ProcessingGraph theoryProbabilityArtificial IntelligenceImage ProcessingData AnalysisData AnalyticsPythonQuantitative ResearchDeep LearningComputer Vision

Education

Georgia Institute of Technology

Master of Science (MS), Computer Science

January 1, 2016 – January 1, 2017

Indian Institute of Technology, Kharagpur

Bachelor of Technology (B.Tech.), Computer Science

January 1, 2010 – January 1, 2014

M.G.M.H.S.S. bokaro

12th, science

January 1, 2007 – January 1, 2009

M.G.M.H.S.S. bokaro

10th

January 1, 1997 – January 1, 2007

Experience

Deccan AI

AI/ML Research

April 1, 2026 – Present

Mountain View, CA · On-site

Sand Hill Angels

Angel Investor

July 1, 2025 – Present

San Francisco Bay Area

Projects

Transitive language translator

October 1, 2017 – Present

The project involves understanding the performance of transitive RNN language translators versus regular RNN language translator using deep learning.

Language Detection using Deep Learning

April 1, 2017 – May 1, 2017

Developed LSTM models for multi class language detection using probabilistic generative model of test string. The decoder LSTM used posterior probability for different languages for prediction. Reported accuracy of 83% for 2 class(English and French)and 60% for 22 classes with different algorithms for generating training sets. Experimented with different methods of training like varied length, fixed length and different conditions on model like early stopping.

GAN for digit image generation

March 1, 2017 – April 1, 2017

Used Generative Adversarial Networks for generating images indistinguishable from MINST digit data. Used keras for developing the deep learning discriminator and generator model.

ML based Quant Market Simulator

January 1, 2017 – Present

Working on NSF funded project on developing market simulator and detect financial manipulation of markets. The work involve designing strategies for AI agents to mimic real world market. The major part of the project is involved in testing for market manipulation using Deep Learning (LSTMs) to learn the behavior of adversarial or manipulative agent.

Reinforcement learner for stock market prediction

November 1, 2016 – December 1, 2016

Developed a QLearner(Reinforcement Learner) which learns the market conditions from economic indicators like volatiity, SMA and Bollinger Bands. Then, based on current indicators, it helps in predicting buy sell or hold action on a particular stock depending on the rewards on various indicators which it has previously learnt.

Personalised movie recommendation system

October 1, 2016 – November 1, 2016

Developed a collaborative filtering based personalised movie recommender system. Given the rating matrix, we tried to extarct the user profile and movie profile usind gradient descent approach and then used the generated user and movie profile to generate ratings for unknown movie and recommend them to user.

Travelling Salesperson Problem

October 1, 2016 – December 1, 2016

Travelling Salesperson Problem is a standard NP complete problem and it can be solved with some approximations using hieuristics. So, I developed 5 different sets of algorithms with different hieuristics to solve it and compare them in terms of time taken and accuracy. The algorithms were branch and bound, simulated annealing local search, MST approx , closest insertion and 2 opt exchange local search. Developed several visualisations to compare the performances of the algorithms.

Stock price predictor based on technical indicators

October 1, 2016 – December 1, 2016

Developed a Machine Learning tool to predict buy sell or hold action on stocks based on various economic indicators like momentum bollinger bands, volaatility and SMA(Simple Moving Average). I used decision trees to predict action on the given stocks based on historical moves of the above indicators. The tool prediction gave nearly 25% profits.

Realistic Topic modelling

September 1, 2016 – October 1, 2016

Developed a system to cluster documents related to same topic. I used bag of words model for documents.The joint probability distribution of words and documents was used for modelling topics. Finally we used expectation maximisation algorithm to calculate latent probabilities and use them to generate top k words related to a topic/cluster.

Localised Real Estate Price Prediction using Stock Market

August 1, 2016 – December 1, 2016

I developed real estate price model using most correlated localised stocks obtained by Dynamic Time Warping algorithm. I used techniques like LSTM and Regression in Python and obtained prediction accuracy within 15% of existing algorithm.I also developed visualisations in D3 to study the dependencies between stock and real estate market.

Distributed file System using Chord Protocol

January 1, 2014 – April 1, 2014

The objective of the project was to maintain a distributed file system with a single view of filesystems for all the users. The files were hashed to different user's machine and the users were a part of a single network.The files were accessed on demand using the famous chord protocol used for distributed hashing.

Sentiment Analysis in cricket

August 1, 2013 – November 1, 2013

It was an academic project on Natural Language Processing.Collected tweets related to cricket for 1 month using Twitter API.The opinion words were extracted and classified as positive , negative or neutral using WordNet and on the basis of the final weights of the opinion words , the sentiment about a particular cricketer was determined. The tools used were Natural Language Toolkit,Stanford NLP Parser ,Wordnet and CMU parser

Virtualisation on mobile phones

May 1, 2013 – July 1, 2013

The project was involved around modifying the existing bootloader of the Operating System to allow using both hypervisor and supervisor mode in the mobile phones at a time. The use case could have revolved around the security issues where we have two Operating Systems on the same device and we want to use secured apps and data in one and the non secured in the other. The project also involved around doing some performance testing to check the overhead coming due to virtualisation and if it is worth having virtualisation on mobile phones after those overheads.

Gait detection using Kinect (B. Tech Project)

January 1, 2013 – April 1, 2014

The objective of the project is to identify the person according to his gait features.Gait is one of the important features that can be used to distinguish individuals from each other and it can be as powerful as a fingerprint.The project consists of creating a database of the gait features of some individuals and try to recognise individuals on the basis of their gait features.The project involved extensive use of image processing , machine learning and data analysis. The project was accomplished in MATLAB and C#. The project helped in decreasing the processing time for gait based identification approach significantly and with good accuracy.

Studying Motif Distributions in Natural Language

January 1, 2013 – April 1, 2013

The objective of the project was to identify the distributions of different types of motif in a natural language corpus and to derive semantic information about different types of motifs present in the corpus.The project was done in Python

Facebook search kind of application

January 1, 2013 – April 1, 2013

A tool which allows user to search anything on the registered users profile .It was similar to facebook graph search tool.The project was accomplished using PHP and MySQL

Online course ranking system

January 1, 2013 – April 1, 2013

The objective of the project was to manage the courses run by an institute and rank the courses on the basis of the users reviews , likes and comments and recommend the most popular courses to the new users.The project was done using Netbeans(Java) and MySQL.

Abandoned Object Detection

August 1, 2012 – November 1, 2012

It was an image processing project with the oblective of indentifying an abandoned object on a busy place by using the video footage of that place and report it as a suspected item. The project was accomplished in MATLAB.

Compiler design for a C type language

August 1, 2012 – November 1, 2012

The project was to design a compiler for a C type language and it was done in C.

KGP RISC

August 1, 2012 – November 1, 2012

The project was to build a 32-bit Single-Cycle Reduced Instruction Set Computer.for a piece of instruction set.It was developed in Verilog.

Transport Company Computerisation software

January 1, 2012 – April 1, 2012

The objective was to automate the order process and its tracking for a transportation company . It also automated the process of truck allocation and scheduling process.The project was using Netbeans(Java) and MySQL

Certifications

SQL

Oracle

June 24, 2026 – Present

Java

Oracle

June 24, 2026 – Present

J2EE

Oracle

June 24, 2026 – Present

Key Strengths

Extensive experience in Machine Learning and Deep Learning, particularly in NLP, Generative AI, and predictive modeling, which are highly relevant to Data Analyst roles involving advanced analytics.
Strong background in data processing, feature engineering, and model deployment from roles at Google and Amazon, indicating practical skills in the data lifecycle.
Experience with various data analysis techniques, including statistical modeling (Gaussian Mixture Models, Monte Carlo simulations), time series analysis (Dynamic Time Warping, LSTMs), and recommendation systems (collaborative filtering).
Demonstrated ability to work with large datasets and optimize data pipelines (e.g., reducing SLA from 10 days to minutes at Amazon).
Academic background from top-tier institutions (Georgia Institute of Technology, IIT Kharagpur) with a Master's in Computer Science.

Cultural & Operational Fit

Cultural Fit Analysis

Soft Skills & Operational Fit