Efficient Team Management: Experience and Life Hacks
Managing projects is a complex task that demands a plethora of skills. The experience of a leader manifests itself even in small details, enabling the more effective organization of team work. In this article, Kirill Meshik, Training Data Team Leader, shares life hacks that can help leaders to work more efficiently within the technological sphere.
What experience in team management do you have?
I gained experience managing teams when I began working as a Project Manager, a role I held for approximately 3.5 years. As a Project Manager, my core responsibilities involved planning and coordinating project execution, distributing tasks among assessors, ensuring their completion, and guaranteeing the fulfillment of client expectations regarding quality and delivery timelines.
For the past 8 months, I’ve been a Team Leader managing a team of project managers. My role revolves around setting task priorities, resource allocation, monitoring goal achievement, handling strategic and organizational communications with clients, overseeing the training and professional growth of new team members, and contributing to the formation and adaptation of operational business processes within the company.
Over time, I’ve gained valuable management experience that aids me in developing collaborative skills, leadership qualities, and achieving results within the projects and tasks assigned to the team.
What are the main tasks and responsibilities of your team in data annotation projects?
Our team works on projects related to data segmentation, detection, classification, transcription, and involves extensive work with LLM (Language Model) and OCR (Optical Character Recognition).
Each project demands significant managerial involvement. Despite common task typologies, we frequently encounter non-trivial and unique cases.
Managers are responsible for piloting projects, dissecting and refining technical documentation, client communication, defending pilot results, as well as training, launching, overseeing, and concluding projects while maintaining performance metrics.
The Team Leader intervenes at critical stages, ensuring process compliance within each project, supporting clients, analyzing dynamic metrics, and adapting or shaping processes based on gained experience and current trends in the field.
What methods do you use for efficient task and resource allocation within the team? What practices do you recommend for effective team management, especially in the field of data annotation?
Our team operates within Scrum processes; all tasks are logged in a task tracker and reflected in roadmaps. When assigning tasks, the Team Leader considers the current capacity of managers and evaluates foreseeable potential tasks. Key evaluation metrics include task complexity scores, the number of assessors in teams, assessors in adaptation periods, and planned/actual time spent.
In managing assessor teams, the manager holds a pivotal position, directly influencing work pace, quality, and team cohesion. To achieve this, several tools are employed: regular meetings, workshops, interactive sessions, shared communication channels, task trackers, team-building activities, among others.
What key performance indicators (KPIs) do you use to assess the success of the team in data annotation projects?
KPIs effectively gauge productivity based on project indicators and personal goals. Aside from assessing current team competencies, a grading system exists to form individual development plans, define long-term goals, and actualize ambitions in work.
Managers monitor progress towards set goals, which serves as an indicator of success for both individual team members and the team collectively.
Have there been situations where the team faced critical deadline failures or unforeseen problems? How did you manage the situation and what lessons did you learn from that experience?
Yes, we’ve faced such situations a couple of times. The critical aspect lies in assessing the possibility of project delays as early as possible. The sooner potential risks are identified, the more tools are available for the manager to solve the problem. It’s crucial to maintain communication with clients, ensuring transparency in processes. This fosters mutual understanding and aids in finding the best solutions.
In data labeling, one of the key challenges is the disparity between pilot data and project data. Even with maximum effort, estimating deadlines based on pilot results might be unreliable when considering project data volumes. In such cases, re-piloting the project with adjusted timelines is the most prudent solution.
Additionally, project timelines might derail due to internal force majeure circumstances. The resolution in such cases is always individualized. This is where a manager’s skills shine: maintaining balance within processes and completely satisfying the client.
How to maintain a balance between delivering high-quality projects and meeting established deadlines?
The quality of outcomes depends on how well the client outlines the technical specifications and how deeply the team immerses themselves in the task. Sometimes, there’s a discrepancy between the client’s specifications and their expectations. In such cases, through collaborative negotiations, we assist the client in understanding precisely what they need.
Another crucial aspect is how the team of assessors perceives these specifications. It’s crucial to avoid misunderstandings. To achieve this, we employ several techniques: absolute clarity in wording, a pool of examples/counterexamples, supplementary materials to elucidate specificities, and individual handling of corner cases. The project manager oversees all these elements. Additionally, our company has a quality control department that aids the manager in promptly assessing dynamics, highlighting problematic areas, and areas requiring intensified attention.
Simultaneously, while planning project resources, it’s essential to adequately assess the timelines for each stage. Every client’s initial request is the same: fast, cost-effective, and high-quality service. During discussions, we aim to showcase our capabilities and orient the client so they can receive a realistic and clear estimation. In practice, there are instances where unusual requests arise: the need for rapid completion, budget reduction, or achieving maximum quality. This invariably leads to the development and implementation of unique solutions.
Understanding client expectations is pivotal for project success because the processes set by the manager are tailored to address specific tasks. Ensuring quality within set timelines primarily involves working with our client’s expectations: understanding them and considering them as the basis for project planning.
Can you describe a particularly challenging project or case that your team worked on during data annotation? How did you approach solving problems, and what strategies did you employ to overcome them?
Currently, our most labor-intensive tasks revolve around LLM-oriented projects. For most clients, these are entirely new areas, so requirements are constantly changing and adapting. In these cases, the manager’s key role is to guide the client based on our experience and align all processes. It might seem surprising, but despite a common typology, approaches can vary significantly.
For example, we had a task where the model’s responses to user queries had to not only consider their context but also determine their mood, crack jokes, incorporate current news, and adhere to a clearly defined style. You can imagine how challenging it is to structure such algorithms and how diverse the training data must be, not to mention their quantity. The project’s timeline was a crucial aspect because such systems can take years to train.
An intriguing solution was found: the automatic generation of training data subsequently subjected to manual moderation, allowing us to reduce dataset formation time by tens of times.
In such tasks, a genuinely creative approach is necessary, often leading to new, unique procedural solutions. The team needs to consist of erudite and creative individuals, professionals in their field. For instance, when a user’s query involves various scientific aspects, the response must be entirely satisfactory, necessitating the competence of the assessor team.
There were even situations when, within extensive team discourse, genuinely complex individual cases were addressed. The primary focus was not just to resolve the encountered corner case but to define the vector for forming responses to such queries within the subsequent annotation.
Were there cases where your team had highly limited resources? If so, how were task priorities determined, and how were resources effectively allocated to achieve project goals?
One basic example is project scaling, i.e., increasing the required resources for a specific manager and their team due to an increase in workload.
My role as a Team Leader involves determining the current capacity of each manager and in foreseeing potential situations. When the resources of one manager are depleted, it’s crucial to promptly involve additional resources in the project. At the same time, it’s essential to establish tangible zones of responsibility for each and monitor the adherence to set goals.
The most crucial aspect here is for the Team Leader to have a strong sense of their team.