Work Corpus
As of 27th April 2025
Raj Patil | raj@cognware.com
Experience Logs
Principal Engineer: ESDS Software Solutions
March 2024 - Present
- Promoted into a more involved and expanded role geared towards optimizing their core product offering : a Data Center Orchestrator
- Researching, designing, and developing cloud computing interfaces across the broader horizontal of Compute, Networking and Storage, along the full vertical of Data Center Orchestration (Hypervisors, DC Operations, Smart Racks, … ) to Cloud Native operationalization (K8S, OKD, OpenShift …)
- Overseeing a growing interdisciplinary research lab that operates across domains : learning whatever the mission demands
- further honing my collaboration acumen
- Also invested in fostering a healthy engineering culture via:
- self-hosting (volunteered for this to further deepen my understanding of the practicalities of DevOps) a somewhat distributed internal developer platform (stitching several open source projects), towards making the lives of internal engineering teams (80 cumulative) easier.
- hosting collaborative internal conferences and talks geared towards keeping the team up to date with SOTA
- one of the technical proxies for inter-institutional (universities, research-labs, industry) engineering collaboration
- evangelizing emacs
AI Engineer: ESDS Software Solutions
Nov 2022 - March 2024
- Developed a comprehensive sentiment analysis portal from the ground up, focusing on customer satisfaction within the ticketing portal.
- now crucial within the organization, used to quantify customer satisfaction and identify areas of improvement with regards to customer experience.
- Headed AI integration across diverse departments, identifying and implementing multi-model pipelines (natural language processing, computer vision and traditional machine learning solutions)
- First bout of architecting and engineering industry-scale AI solutions: the most prominent being School Education and Sports Department of Maharashtra winning a Guinness World Record, for which ESDS was the technical facilitator.
- Mentored individuals across domains and began honing my managerial/leadership aptitude.
Deliberate Hiatus
Aug 2022 - Nov 2022
- Deliberate career break to realign my personal, career, health, and learning objectives
- Overall health had taken a hit and needed to take a step back : later realized major underlying factor to be severe deficiencies in essential micronutrients
Goldman Sachs: Analyst
July 2022 - Aug 2022
- Briefly worked as a Strat in the Controllers Department, gaining valuable exposure to the financial industry.
- Although a shorter engagement, gained insights into the organization and its operations.
- Took a career break to address serious health concerns
IIT Hyderabad | KLA Tencor: Research Assistant
Aug 2021 - May 2022
- Collaborated with Prof. Vineeth N.B. on research to develop a generic transferability score for inter-domain transfer-learning tasks.
- Focused on computer vision applications within chip technologies.
- Leveraged knowledge in Domain Adaptation (Transfer Learning) to address relevant challenges.
Goldman Sachs: Summer Analyst
June 2021 - July 2021
- Worked towards enhancing (unsupervised) clustering interpretability.
- Developed a versatile multidimensional hyperparameter grid search framework to automate model optimization and fine-tuning.
- Was a remote engagement due to COVID-19 : valuable experience working with a distributed team
CVPR 2021 Workshop paper | AUTONUE 2021 : Research Assistant
- Part of the organizing team for a domain adaptation challenge for semantic segmentation, contributing to a workshop paper.
- baselines available here : https://github.com/rajp152k/AUTONUE-III-Baselines
- Guided and mentored by Prof. Vineeth N.B. and Prof. C.V. Jawahar
- This is when I began using reverse chronological streams of work logs to serve as indices into progress
Projects
| Project | Synopsis | Principal Domains |
|---|---|---|
| Nth-Roam | An Emacs Package to efficiently manage multiple instances of emacs org-roam | Software Engineering |
| fabric-gpt.el | An Emacs Package to employ crowd sourced prompts into an llm-client intuitively | Software Engineering |
| Cloud Native LLM inference servers on heterogeneous compute infrastructure | Horizontally scaled self hosted open source LLMs (OpenLLM, Kubernetes, nvidia-container-toolkit, VMWare ESXI) for serving LLMs (vLLMs) to the engineering and support teams on heterogeneous compute infrastructure (Nvidia(T4s & H100s), AMD (MI300Xs and MI300A), X86 and ARM) | (LLM:ML)Ops |
| LLM Augmented Continuous Integration with Self Hosted Giltab | Kubernetes to manage varied workers along with SOTA development and integration practices (Linting, Static and Dynamic Application Security Testing (SonarQube) along with LLM augmented security analysis (Ollama, Langchain, LlamaIndex)) | NLP, Dev(LLM:ML)Ops, CyberSecurity |
| Cloud Native Web Application Firewall | Scalably productized(Kubernetes Horizontal Pod AutoScaler & Event Driven AutoScaler (KEDA : CNCF Grad)) and benchmarked(Apache bench) Coraza(an OWASP compliant go WAF library) | CyberSecurity |
| CLI timer | A bash func to save your soul | CLI-fu |
| Storage Virtual Machine DR Replication Monitoring | Designed, developed and operationalized a golang microservice cluster to manage real time analytics streaming to be integrated into the existing observability stack (NetApp ONTAP, RabbitMQ, InfluxDB, Prometheus) | Real Time Observability |
| Guinness world Record : Deduplication, Anomaly Detection, Tampering Detection | context elaborated in blog link | Computer Vision |
| Sentiment Analysis for Support Tickets | implemented a primal ETL, fine-tuned a Language Model (weakly-supervised) over a (pseudo-labelled) stratified subset, inference jobs formed integral part of business analysis crons | Natural Language Processing, Data Engineering |
| Quantum Error Correction Codes | created a framework to test efficacy of qubit allocation strategies for noise-resistant quantum computation | Quantum Computing |
| Tic-Tac-Toe Reinforcement Learning | tested out RL algorithms with random and heuristics based agents | Reinforcement Learning |
| Option Pricing via Monte Carlo Simulations | explored Monte Carlo option pricers, while comparing results with conventional stochastic calculus models (Black Scholes) | Stochastic Calculus |
| Secure Multiparty Computation (for Deep Learning) | Survey on how cryptographic techniques are used to maintain secrecy during multiparty computation (neural networks in this case) | Cryptography |
| WDGAF(We Do Give a Figure) | Created plotting DSL using principals of compilers and published it to pip | Compilers |
| Relational Databases and Data Hygeine | Executed complex relational database queries on the imdb movie database; implemented scraping scripts to fetch relevant data | Relational Databases |
| Optimizing scheduling algorithms | researched into tweaks to optimize the scheduling algorithms BOCC, BTO, MVTO | Distributed Computing |
| Musical Chairs: inter-process coms | Simulating a game of musical chairs using modern concurrency constructs from C++11 onwards | Concurrency |
| Chat Room App | created a command line chat room app using principles of socket programming | Computer Networks |
| Custom Deep Learning Framework | created a deep learning library from scratch (axioms being high school math ) | Software Engineering |
| Mind Muscle Matrix | My first blog. I was fresh into fitness, had watched the matrix more than a dozen of times and was overdosing with some heavy philosophical works. Setup a VPS, wordpress, mailing and distribution pipelines. | Business |
| Calculator | wrote C to simulate the finite state machine on an ATmega16U2 (Arduino), designed and soldered the circuit (buttons, display, and the ATMega) on a veroboard, functional with add,mul,sub,div | Electronics |
| 3D-printed pin ball machine | first major project at IIT Hyderabad. designed the individual components from scratch in CAD and assembled the same (star wars themed - had a millenium falcon, a slide, a star trooper helmet and it came in black) into a functional pin ball setup | Computer Aided Design, Digital Fabrication |
| PLC HMI Design | first explicit experience (in hindsight) with Finite State Machines when I was 14 as an Intern at Nash Robotics. analysed and reasoned about the states and corresponding control flows on programmable logic controllers. | Digital System Design |
Extracurricular
Teaching Assistance
- Prof. Vineeth N.B. - CS5590 : Foundations of Machine Learning
- Prof Srijith P.K. - AI1100 : Intro to Artificial Intelligence
FitSoc 2021-2022
- Established a fitness community dedicated to supporting college students in balancing their fitness and academic pursuits through a hybrid athleticism approach.
- Organically bootstrapped and grew the community, currently comprising over 200 active members.
Founding Core Member Epoch (The ML club of IIT Hyderabad) - 2019
- Conducted the inaugural session of the club.
- The session centered on exploratory data analysis utilizing key tools and libraries, including Pandas, NumPy, SciPy and matplotlib.
National Service Scheme
- participated in several cleanliness and plantation drives during my time at IIT-Hyderabad
- created several social awareness posters regarding various societal issues
Conferences Attended
This should serve as a quick index into my networking interests during phases of my career
| Conf | When | Where | My Focus and Incentives |
|---|---|---|---|
| NVIDIA AI Summit | October 24-25 2024 | Mumbai, India | Successful LLMs in Prod, Networking |
Education
B.Tech: Indian Institute of Technology Hyderabad (2018-2022)
All India Rank (JEE Advance 2018) - 658
- Major (2018-2022) : Computer Science and Engineering : 9.26/10
- Minor (2020-2022) : Artificial Intelligence : 9.50/10
AutoDidacticism (2000-Present)
- reading, writing, thinking, conversations, teaching, disciplined curiosity and making mistakes play an essential role in how my learning trajectories unroll..
- I’m not trying to build a career any more than I’m trying to explore the world epistemologically, in general
- This was an index into major deliverables that might be societally gauge-able
- As for my generic incentives, this recursive-introspective terser composition is a summarization attempt