Vinayak Chaturvedi

In my current role at Syndigo LLC, I built robust AI and ML infrastructure from the ground up. My journey has
been marked by impactful achievements:
 Scaling Success: I’ve led large-scale ML projects within a tight-knit team of four, demonstrating ownership and
expertise. From inception to deployment, I’ve navigated the entire project lifecycle.
 Problem Solver Extraordinaire: My toolkit includes data analysis, model development, ablation studies,
optimization, fine-tuning, RESTful API development and MLOps. I’ve turned complex challenges into elegant
solutions.
 Strategic Orchestrator: Collaborating with key stakeholders and leaders, I’ve orchestrated requirements gathering
sessions and brainstorming marathons. My strategic project planning ensures alignment with business goals.
 Cutting-Edge Research: The ML landscape is ever evolving, and I thrive on staying ahead. I’ve contributed to
applied research, bringing state-of-the-art solutions to the table—always mindful of practicality and relevance.
 Mentorship and Knowledge Sharing: Guiding junior team members is my passion. I’ve hosted year-long companywide ML training sessions, fostering a culture of continuous learning.
 LLM, NLP and Computer Vision Champion: I’ve led successful projects in Large Language Models (LLM), natural
language processing (NLP), speech synthesis and computer vision. Transforming data into insights is where I excel.
 Seamless Integration: Whether it’s securing models or integrating ML services into existing tech stacks, I ensure
smooth transitions.
Worked on wide variety of target tasks such as classification, summarization, question answering and generation
on textual, audio, image/video data. Conducted and led applied research on various ML techniques/models to solve
the underlying business problems generating significant financial revenue and immense end-user satisfaction.

My experience is resonant but not limited to:
1. Computer Vision:
o Feature-Based Methods:
i. Proficient in SIFT, SURF, and ORB—classic techniques for feature extraction and matching.
ii. Leveraged these methods for tasks such as object recognition, image stitching, and scene
reconstruction.
o Siamese Networks:
i. Designed and trained Siamese neural networks for similarity learning.
ii. Applications include face verification, product verification, and tracking objects across frames.
o CNN (Convolutional Neural Networks): Extensive experience with CNN architectures for:
i. Image classification (e.g., ResNet, VGG, Inception).
ii. Object detection (e.g., YOLO, Faster R-CNN).
iii. Semantic segmentation (e.g., U-Net, DeepLab).
o Large Language Models/ Diffusion:
i. Achieved high quality image in-painting and out-painting using diffusion models like stable
diffusion (SDXL).
ii. Built pipelines to integrate GPT-Vision into products.

2. Audio and Multi-Modal Models:
o Audio Deep Learning:
i. Investigated WaveNet, MelGAN, and Tacotron for speech synthesis.
ii. Applied CRNNs for music genre classification.
o Multi-Modal Fusion:
i. Explored late fusion (concatenation) and early fusion (parallel processing).
ii. Integrated vision and audio features for tasks like scene recognition within videos.
3. Natural Language Processing (NLP):
o RNNs (Recurrent Neural Networks):
i. Developed and fine-tuned LSTMs and GRUs for sequential data modeling.
ii. Applied them to tasks such as sentiment analysis, language modeling, and time series prediction.
o Entity Matching:
i. Devised algorithms for entity resolution across large text corpora.
ii. Utilized techniques like TF-IDF, Jaccard similarity, and Levenshtein distance.
o Textual Entailment:
i. Constructed models to determine logical entailment between sentences.
ii. Incorporated attention mechanisms and word embeddings for context understanding.
o BERT Models:
i. Fine-tuned pre-trained BERT models for specific NLP tasks.
ii. Achieved state-of-the-art results in sentiment analysis and question answering.
o LLMs (Large Language Models)/Transformers:
i. Explored transformer architectures (e.g., GPT, T5) for various NLP applications.
ii. Implemented self-attention and multi-head attention layers.
4. ML Techniques:
o Quantization and Pruning: Optimized neural networks during deployment by reducing weight precision
(quantization) using LORA and QLORA and removing unnecessary connections (pruning) using PEFT.
Achieved a delicate balance between accuracy and efficiency.
o Neural Architecture Search (NAS): Achieved a delicate balance between accuracy and efficiency using
Vertex AI (Google). This required a lot of computation resources and budget.
o Core Frameworks: Proficient in both PyTorch and TensorFlow frameworks. Designed custom layers,
implemented complex loss functions, and appreciated their unique features.
o Hardware Optimization: Explored parallelism, memory hierarchy, and energy efficiency trade-offs
amongst various CPUs and GPUs, deploying models on relevant hardware. LPU’s and NPU’s are popular
alternatives but our requirements did not have a necessity on that hardware.
o Hyperparameter Tuning: Fine-tuned models using grid search, random search, and Bayesian
optimization. Strived for optimal accuracy-efficiency trade-offs.
o Knowledge Distillation: Compressed large models while preserving performance. Explored techniques
like temperature scaling and attention transfer to achieve compressed student neural networks.
5. Production grade software systems for AI:
o End-to-End Pipelines:
i. Architected robust pipelines handling data prep, training, and deployment.
ii. Ensured seamless integration into micro-services with version control, monitoring, logging and
securing.
iii. Utilized Azure Kubernetes clusters to create the production host environment.
o RESTful APIs:
i. Designed RESTful APIs enabling CRUD operations using Python, FastAPI and flask.
ii. Generated requirements and guidance for software developers on MLOps and DevOps.

• Developed a MATLAB Application for National Road Research Alliance and Minnesota Department of Transportation with contract number : 1034192 (Research Grant).
• Link : https://dot.state.mn.us/mnroad/nrra/structure-teams/geotechnical/files/1034192-load-restriction-task-5-deliverable.pdf
• Designed, implemented and deployed an application (frontend and backend) based on MATLAB that predicts a safe passage on asphalt roads for heavy duty vehicles after a rainfall event.
• Successful presentation and demonstration to Minnesota Department of Transportation.
• Reviewed all documentation for accuracy, quality and compliance.
• Took corrective action to restore or maintain quality standards.

• Provided computer programming subject matter expertise to undergraduate students in C, C++, Java, Python and Javascript.
• Developed an application for user logging and scheduling at programming and assistance centre.

• Taught computer science undergraduate 400 level courses for over 30 students.
• Checked assignments, proctored tests, and provided grades according to university standards.
• Mentored students through office hours and one-on-one communication in assignments, labs and class projects.
• Coordinated homework and assignments for absent students.
• Worked with teachers in evaluating student progress, needs and gains.

• Planned lessons for allotted time to strengthen weak subjects and build programming skills.
• Collaborated with students to complete projects and assignments, identify lagging skills and correct weaknesses.
• Made lessons interesting and engaging using art and visual aids to bolster learning.
• Maintained detailed files and created charts tracking students learning progress.

• Supported classroom activities, including tutoring, grading homework and reviewing exams.
• Partnered with teacher to plan and implement lessons following school's curriculum, goals and objectives.

• Performed IT Audits for 30+ local and multi-national companies across multiple sectors like Banking and Capital Markets, Manufacturing and Oil and Gas.
• Performed Python code reviews and built VBA tools for automating jobs and improving project efficiency.
• Professional experience with enterprise resource management tools like SAP and Oracle ERP.
• Provided recommendations on board of directors and stakeholders on group's effectiveness, actions and future plans.

• Provided front-end website development using WordPress, Hubspot and other editing software.
• Planned website development, converting designs into usable web presence with SQL,HTML, PHP, JavaScript, CSS, AJAX and JSON coding.
• Employed search engine optimization tactics to increase reach of targeted audience.
• Oversaw technical issues and troubleshooting requests to resolve surfaced problems.

• Developed user interfaces with modern JavaScript frameworks, HTML and CSS.
• Designed and implemented web components across new and existing designs.
• Participated in project development through entire Software Development Lifecycle (SDLC).

• Received Graduate Assistant Scholarship.
• Received scholarship from National Science Foundation (NSF) for attending annual National Robotics Initiative (NRI) conference at Washington, D.C. (an invitation only meeting).