Authenticated Chat Experience

Travis' first project at Slalom was a multifaceted Chatbot initiative. He was tasked with integrating high-priority features from the backlog, and ensuring solutions were feasible with input from Solution and Technical Architects. His role extended to scripting for data seeding, refactoring code to meet new API requirements, and facilitating the bot's mobility across Salesforce organizations. Additionally, Travis led the effort to design and document the onboarding process for new agencies and developed authentication mechanisms for personalized user interactions. This project laid the groundwork for future phases aimed at expanding bot and Live Agent functionalities to other agencies.

• Worked closely with Solution Architect to ensure the viability of proposed solutions and provided Level of Effort estimates for work prioritization and scheduling.
• Developed a script to seed complex data structures utilizing advanced import/export capabilities of SFDX (Now SF).
• Refactored the chatbot and associated Apex code to align with updated API requirements and facilitate bot deployments.
• Enhanced bot functionality to allow seamless transfer between different Salesforce organizations.
• Identified and seized opportunities to mentor junior team members, fostering their growth in areas of expressed interest.
• Lead the development and documentation of a comprehensive onboarding plan for new agencies to the existing bot and Live Agent setup.
• Authored detailed deployment instructions for the chatbot tailored to new agency integration, laying the groundwork for upcoming projects.
• Engineered and implemented a system that uses ForgeRock authentication tokens, enabling the chatbot to deliver personalized user experiences. This involved close collaboration with ForgeRock and the client team, resulting in a bot that adapts to user login state and modifies the experience accordingly.

CI/CD Updates (Github)

The Global Digital Services team wanted to update repository structure and introduce automation to more effectively manage internal assets in an effort to reduce the learning curve and friction involed with deploying assets from the library for less technical team members.

The primary challenge was configuring the pipeline to handle a variety of metadata subsets consistently. This range included an Experience Cloud site, a chatbot, various code segments, their associated documentation, tests, and deployment scripts.

Leading two developers, Travis and team successfully developed a robust pipeline with practical quality checks, which stands out for its sophistication in the context of Salesforce projects allowing the deployment of individual assets and featuring scripts capable of populating a new development environment with data while preserving relationships within a given data-set.

He also introduced the concept of tokenization for metadata that typically varies during the deployment process as it moves through the pipeline, which is especially relevant when deploying community sites, embedded services and the like.

Furthermore, the continuous integration (CI) process now automatically manages a series of activations and configurations, eliminating the occorance of developers forgetting to activate things after deployment.

• Guided a collaborative team effort, enhancing skills in Salesforce data exports with relationship preservation and BitBucket pipeline automation.
• Designed a new branching structure that not only supported seamless pipeline automation, but accommodated forking for client-tailored adjustments to assets.
• Set up new Salesforce environments for Development, Integration, and Stable, with the Stable environment dedicated to asset and capability demonstration.
• Integrated quality gates within the pipeline to enhance code quality using Apex PMD.
• Implemented JWT authentication to Salesforce using SFDX and environment secrets, reducing script duplication.
• Updated Apex classes to align with new quality gates and testing requirements.
• Created seed scripts to efficiently populate new development environments with necessary data.
• Tokenized metadata that changes between orgs, such as admin emails on experience sites.

Statewide Agency Bot Rollout

Following the initial delivery of the Chatbot to the client, Travis was engaged to design and develop tailored solutions for individual agencies as they were onboarded to the existing LiveAgent implementation and Einstein Bot. His involvement was on an as-needed basis, requiring him to rapidly acclimate to the project's evolving context to ensure the delivery of quality solutions in a timely manner

• Set up the necessary scripts and CMDT structures for bot deployment across new sites.
• Resolved arising issues to maintain consistent bot performance.
• Offered technical advice to the team when needed.
• Updated the bot's embedded script to trigger new behaviours based on certain search parameters.
• Provided one-on-one guidance to the client's developer for a clear understanding of the bot's implementation.
• Improved the test site by developing abstractions, reducing the time required to add new agencies.

DE Team Branded Demo | Asset Library

Travis took on a personal project to improve client interactions by integrating Lightning-Out and Salesforce APIs. He single-handedly created a website with GatsbyJS, Tailwind CSS, and Hygraph, demonstrating the potential for seamless integration of Salesforce solutions into a client's existing technological landscape. The platform was designed to simplify the process for business development teams to present customized, brand-aligned demos, aimed at strengthening client trust.

This website allowed for real-time customization of demonstrations, enabling business development to adapt presentations to match a client's branding instantly. Travis also introduced an 'asset library shopping' experience, giving clients an active role in tailoring the demo to their preferences, which showcased a dedicated approach to engaging and innovative client service.

• Setup and configured Hygraph as a headless CMS and central location for application configuration.
• Built website using Gatsby (built on React), Hygraph (formerly GraphCMS) and TailwindCSS.
• Deployed site to S3 bucket.
• Wrote and configured integrations between AWS, Salesforce, and Hygraph.
• Setup use-case demonstration using Salesforce Einstine chatbot and a dummy website showcasing the impact of having a means of quickly branding an existing asset to match client websites.

Salesforce | Off Platform Pub/Sub (Avoiding Salesforce Entitlements)

A large auto rental company sought to reduce the 'swivel chair' problem in their call center, where agents were required to reference information across three screens from several external applications. Travis' team (part of a comprehensive program) was tasked with the integration of Salesforce with RingCentral. The business objective was to enable agents to swiftly confirm the identity of callers by retrieving information from the Salesforce backend utilizing data provided by RingCentral.

Travis initially proposed leveraging the Salesforce CRUD API to insert records and trigger platform events. This approach would log interactions for compliance purposes and broadcast a system-wide message via the Salesforce event bus (Platform Events). The client, however, expressed concerns about unexpected costs related to Salesforce platform entitlements.

Consequently, Travis suggested an alternative off-platform solution leveraging AWS IoT Core, which brough with it the added benefit of facilitating easier and more economical integration of subscribers from various platforms.

• Designed and proposed a solution leveraging the Salesforce CRUD API and Salesforce Platform Events, which was not pursued due to the associated costs of platform event entitlements.
• Devised and recommended an alternative off-platform publish/subscribe model utilizing the MQTT protocol, backed by AWS IoT Core.
• Selected AWS as the service provider given the client's existing relationship with AWS and the fact their usage would consistently remain within the free tier, avoiding additional costs.
• Prepared comprehensive documentation for both the preferred and alternative solutions, including a less conventional fallback option for the client's Technical Advisory review board.
• Authored the MQTT client using the paho-mqtt library due to its compliance with the Salesforce Lightning Locker Service and its permissible licensing for commercial application. If the client had been using Lightning Web Security, the AWS SDK would have worked.
• Integrated the MQTT client with the clients service console to implement desired UI.
• Provided detailed technical documentation describing solution architecture and operation.

Leading AI Solution Integration & Training for Client Success

This engagement was unique in Travis' experience, as the client was more interested in educating themselves and understading the value proposition given some painpoints they wanted to discuss as possible candidates.

He led the initiative by providing expert insights on Generative AI, offering specialized training, and showcasing its practical applications through the development of a Proof of Concept (PoC). Although not ready for production due to the eight-week timeframe, the PoC was comprehensive and required only minor adjustments to meet peer review standards and be promoted to production.

The PoC, dubbed Generalized Object Summarization, is an easily configurable extension to the Slalom Generative AI Framework for Salesforce. With minimal effort, any admin capable of writing a SOQL query can declaratively construct a prompt similar to how email templating is done.

The PoC proved to be a key educational tool offering something concrete and relevant that participants could engage with and understand. Training started with an in-depth look at Retrieval Augmented Generation (RAG) and the system-level architecture of our framework, followed by practical sessions on utilizing the PoC. Once participants grasped the overarching thought process, we delved into prompt engineering, concentrating on Chain of Thought (CoT) and Skeleton of Thought (CoT) patterns.

The work Travis did on this project is currently being sold into the market as part of Slalom's Generative AI Framework for Salesforce.

• Designed, developed, and deployed the Generalized Object Summarization feature for the Slalom Generative AI Framework, adding the ability for any admin capable of writing a SOQL query to declaratively construct a prompt similar to how email templating is done.

• Authored comprehensive internal documentation and successfully advanced the feature into our production Asset Library following a rigorous peer review.

• Created and delivered detailed documentation on the application's architecture and operational procedures for client reference and use including some ideas on how to extend it to gain additional value.
• Conducted in-depth discovery interviews with diverse company stakeholders to identify potential future use cases and to gain insight into their requirements, expectations, and concerns.
• Facilitated weekly training sessions for the client's technical team members, fostering knowledge transfer and skill development in Generative AI.
• Presented new framework feature to sales team to explain what it is, how it works, and where it can be used. Fielded questions and provided documentattion as needed.

Authenticated, NLP Enabled Transactional Chat

A large provider of broadband internet services wanted to get started with NLP enabled transactional chat bots (Einstein) and an authenticated user experience allowing their users to conduct account level business through chat without having to interact with a human effectively improving deflection of calls/chats routed to humans in a call-center.

• Integrated wtih GTM objects to get privelaged user information needed for identifiying the user in Salesforce and as a flag to know they are authenticated.
• Designed and built an application allowing an admin to confgure a UI form element similar to a details page or highlights panel by selecting data and identifying where it shoud be placed. This is useful in circumstances when the data you need on page is reachable by relationship, but not very close, or requires multiple queries.
• Added to internal Asset Library after passing requested peer review.

Realtime Feature Adoption Monitoring

As a rapidly growing SaaS company, PI (The Predictive Index) wanted to track feature adoption by inspection of user interaction with the platform. To meet this requirement, Travis designed and built a slack integration that could be easily implemented anywhere it was needed to identify the use of a particular feature. This was made easier with the use of feature flags. This project took place before Salesforce bought Slack, so the canned integration tools didn't exist yet.

• Designed and built defensive callout logic so the same code could run during async and sync operations. This reduced the amount of code that needed to be written and maintained.
• Designed with scalability in mind. Built abstractions for easy reimplementation.
• Worked with stakeholders to refine functionality as continued use lead to insights not thought of during ideation.

New Materials Ordering System

Part of the product offering at PI is training materials for consultants to use as they carry out their business with clients. Order fulfillment wanted a new ordering system that reduced friction and unified the different parts of the fulfillment process. To that end, Travis built a new front-end and back-end collaborating with a UX expert and order fulfillment to meet their needs.

• Worked with UI/UX architect providing guidance that helped balance the ideal state and a state that could be itterated on easily following an initial release.
• Implemented front-end following design specifications set forth by the UI/UX architect.
• Worked with the fulfillment department to understand and implement scalable back-end processes.
• Baked administrative abstraction into the solution where it made sense to empower administrative users to reconfigure things as needed.
• Used Declarative process automation when it was likely business process would change to empower administrative power users to make changes where business logic was not to complicated.

New DevOps Program

One of the topics discussed during his interview process was the desire for more structure on the dev team. Travis had experience with the kinds of tools the team would need and once onboarded began advocating for those tools and processes. He ultimatly got buyin from the Architect who secured funding and partnered with him to select and implement a tool-chain everyone including admins could use. A credit to the team he was working with, adoption was high to start, and while there were slips and misses here and there, the improvements in code/config quality was evident in the newfound stability prodcution was experiencing.

• Worked with vendor (Copado) to design and implement their tools.
• Configured Copado with quality gates and process automation to reduce friction.
• Developed and documented formalized process with buyin from department head and team.
• Implemented and managed a regular production release cycle.
• Implemented and lead regular (no judgement) design/code reviews leading to significant reduction in production bugs.
• Created and executed transition plan to ensure adoption of process and tools would be successful.

CAD/AVL Platform Refresh

The CAD/AVL technology stack was aging and very heavily dependent on a single vendor. In an effort to update the entire technology stack for the fleet and reduce dependency on a single vendor, an RFP was issued to find vendors that better aligned with long-term technology goals without an engineered vendor-lock. The proposal would need to improve upon vehicle telemetry reported in real-time, frequency-based scheduling as well as arrival-based scheduling, and make use of hardware already deployed where possible as well as a mobile app for improved rider experience.

• Documented existing platform capabilities and worked with many different departments to document business requirements for the purpose of issuing a RFP.
• Evaluated proposed solutions from all bidding vendors utilizing a weighted scoring system from initial call for proposals through down selection and eventually final vendor selection.
• Served as technical lead for the project, working with all business areas and vendors to translate business needs into technical requirements while making sure conflicts and technical dependencies were resolved quickly and effectively.
• Reverse-engineered proprietary sign protocol (a modified version of an open protocol called Alpha) by inspection of raw hex data captured directly from the vehicle area communication network (SAE J1708). Documented specification for protocol and provided to the vendor resolving an integration issue that would have cost hundreds of thousands of dollars in new hardware.
• Assessed functionality against requirements during UAT while following vendor script. This presented a challenge because the vendor script did not touch all requirements. Documented delta between requirements and demonstrated functionality as well as usability enhancements to be completed. Presented findings and Go/No-Go Recommendation to Vice President of Technology.
• Worked with vendor engineers to provide information needed to build the solution.

Google Maps Integration

As part of their strategic effort to modernize ridership experience, Indygo wanted to publish realtime telemetry data to Google maps but had been unable to gain approval to have it published to the general public. Travis was tasked with identifying the root cause of, and remediating issues preventing publication of real-time telementry data to the public by bringing source data into compliance with Google's standards. When his work was complete, ridership was able to see where their bus was in real-time, understand if it was running early, late, or on-time, and by how many minues it was running early or late.

• Assessed and documented the root cause of data issues preventing IndyGo from going live on Google Transit.
• Performed a gap analysis and documented an approach to resolving issues identified.
• Wrote .Net utility to groom schedule data into the format required to align proto-buffers provided by AVL platform with Google data quality standards resulting in an acceptance of real-time data to Google and eventually publication of telemetry data to Google Maps.
• Chased down rolling stock hardware issues representing gaps in real-time telemetry data and remediated.
• Worked with Google to ensure alignment of data and publication of data to Google Transit for public consumption.

RFID Reader Remediation

Having just opened a new transit center (the Julia M. Carson Transit Center) in the heart of Indianapolis, IndyGo was experiencing some unexpected behaviours related to the real-time arrival and departure data presented on signage located at each bay where riders could look up and see the next bus to arrive and when it waas expected to arrive and depart.

Occationally, two or more bays would swap information requiring additional staff to monitor for the behavior and make sure riders got where they needed to go. Travis was tasked with root-causing the problem, formalizing his findings, and formulating a plan for remediation. Having done so, he led a team of 3 and coordinated the activities of 2 contractors to implement his remediation plan which successfully resolved the issue.

• Performed root-cause analysis of why real-time signage would often swap information with one another. Efforts included physical and electrical inspection of RFID readers and real-time signage, inspection of vendor source code and configuration files, and proving root-cause hypothesis through testing, then reliable and repeatable demonstration of the problem on-demand.
• Formally documented findings and recommended remediation plan.
• Presented findings and remediation plan to executive leadership successfully gaining approval to execute on said plan.
• Executed recommended remediation plan managing the activities of 3 other people over the course of several weeks. This required coordination of vendor activities and internal resources around scheduled arrivals at the downtown transit center to minimize impact to operations.

OTA Update Process Optimization

OTA (Over the Air) updates were problematic as they had to be done while the vehicle was in the garage, turned on, and given time to boot in order to take an update to schedule data correctly. This process took longer than it did for the average operator to turn the vehicle on and leave the garage. For this reason, when an update included schedule information for the MDT (Mobile Data Terminal), and for whatever reason a coach operator did not wait for the update to take and the MDT to restart their rout information including on-time performace would not function as it would not have valid rout infromation for the current date. This would force the operator to interact with dispatch much more frequently and if they did not have the route memorized on-time performance would suffer significantly. Travis saw this problem, outlined a solution, worked to get buyin from the two other departments required to formalize a new process and owned the execution of that process moving forward.

• Outlined and proposed initial process including communication plan.
• Worked with planning and maintenance departments to drive concenceus through thoughtful discussion taking into account differing and at times conflicting interestest.
• Produced formal documentation for agreed upon process and managed its execution in concert with leadership from other groups involved.

Signed Long Integer to IPv4 Converter

EEPROMs used to provide configuration to the computers on rolling stock would sometimes need updates to network information as the network topology sometimes changed. IP addresses were stored as signed long integers instead of dotted IP4 addresses. Doing the conversion by hand is straigtforward, but prone to human error. Travis saw an opportunity to reduce human error and improve productivity by writing a .Net utility that would do this conversion for him in both directions.

• Wrote .Net utility that can convert a signed long integer to a dotted IPv4 address and back.
• Tested to ensure reliable operation.

Mass EEPROM Binary Update Utility

Vendor provided EEPROM configuration tools only worked on one EEPROM at a time. When needed, it was not possible to perform a mass update to parameters common to the entire fleet. This was a time consuming and tedious process prone to human error and an exellent opportunity to add automation to the process.

• Identified EEPROM binary variable locations by inspection of hex.
• Wrote .Net utility that could take a reference EEPROM binary and deploy the delta across all versions in a remote filesystem using knowledge gathered during inspection.
• Tested to ensure reliability of solution.
• Documented tool usage.