Confirmed on the employer's own hiring board on Nov 17, 2026. First seen by Alion on Nov 17, 2026. Solen Software Group scores B on the Alion truth index.
Job Title: Backend Engineer, Data & Integrations
Company: Solen Software Group
Location: Karachi/Lahore, Pakistan
Employment Type: Full-time
About Solen
Solen Software Group is a long-term owner and operator of mission-critical B2B software companies. We acquire proven, profitable businesses and partner with their teams to build lasting market leaders. Unlike traditional private equity firms, Solen invests with permanent capital: we buy to keep, not to sell.
About the role
As our AI Engineer, you will own the AI platform used by our research, M&A and operations teams and by the businesses we acquire: the overall product and the AI workflows that run on it.
The role has broad technical scope over time, from AI agents and the memory layer they use to the interface the team works in every day. You will start focused: one useful research workflow, built on real sources, with a clear way to tell whether its results are right. You will work closely with business teams to turn how they make decisions into reliable tools.
Key responsibilities
First-year priorities
- Deliver a research workflow that the team uses every week, working with real sources and citing them in every output.
- Build a way to evaluate that workflow's results against the team's own work, and use it to decide what to improve next.
- Extend the same approach to our M&A workflows, connecting our existing skills into agents that hand work to each other.
- Put access by role and accuracy tracking in place from the start.
Product and architecture
- Own the platform as a product: its architecture, its roadmap and how the team uses it, starting with research and M&A and extending to the companies we acquire.
- Evaluate and extend existing tools, both public or open-source and Solen-internal, where they fit, and build new components where they do not.
- Design so the platform gets better as AI models improve, rather than building around gaps the next model release will close.
- Send each task to the right model, for example frontier models for complex reasoning and smaller or open-weight models for routine, high-volume work, balancing accuracy, cost and speed.
AI workflows and agents
- Build agents for research, M&A and operations that share context and hand work to each other.
- Turn the team's decision criteria and playbooks into versioned, testable skills kept in one shared library
- Build retrieval and structured extraction from documents, call transcripts, emails and web content, with sources cited.
- Build a decision log that records each decision and what happened next, so the system learns from results over time.
- Later, build an assistant for the CEOs of the companies we acquire, designed once and configured for each new company.
Interface and shared knowledge
- Build the interface where every team member works with their agents, with permissions by role and connected to the tools we already use, such as Salesforce, Slack and Microsoft 365.
- Build a shared, searchable memory of meetings, documents and decisions that gives more weight to recent information and works beyond a single model's context limits.
Evaluation and governance
- Build evaluation into every AI workflow: test sets of real tasks with expected answers, automated and model-graded checks, tracing and monitoring as models change, human review rates and error logs.
- Set rules for what content is stored and remove sensitive information, such as compensation, before it is used.
- Keep our data and memory under our control, separate from any single AI provider, and review the security of AI vendors.
- Set AI usage budgets by team and task, track spending against them and add alerts.
Team and adoption
- Map workflows with business users, automate the most repetitive ones first, and drive adoption.
- Document systems so business users can maintain their own workflows.
Key requirements
- Bachelor's degree in Computer Science, Engineering, Data Science, Mathematics or a related field; a Master's degree is a plus.
- 4 or more years building and deploying software, data or AI products in production, including at least 2 years building LLM applications (prompt design, retrieval, agents, structured extraction and evaluation).
- Strong Python and SQL.
- Hands-on experience with LLM APIs (for example Anthropic or OpenAI), retrieval and agent frameworks, and a clear view of when not to use them.
- Experience evaluating LLM outputs with test sets and automated checks.
- An AI-first way of working: you use coding agents such as Claude Code or Codex every day and can show how they speed up your work.
- Experience building web applications or internal tools that non-technical users rely on (for example React or Streamlit).
- Working knowledge of a cloud platform (AWS, GCP or Azure), Git and data security practices.
- Enough financial literacy to follow a deal model and read a P&L, balance sheet and cash flow statement.
- Clear written and spoken English, and the ability to explain technical trade-offs to non-technical colleagues.
- Willingness to commit to the role for at least three years.
Nice to have
- Experience designing memory, context or permission layers for LLM applications.
- Experience with MCP servers, model routing, fine-tuning or open-weight models.
- Product management experience, or experience owning an internal product end to end.
- Experience in financial services, private equity, consulting or corporate development.
- Experience as an early technical hire, or hiring and leading engineers.
What will set you apart
We are looking for a builder who likes open-ended problems and wants to own a product for years, not ship a single project. Technical depth is essential, but how you work with people matters just as much.
- Builder and owner: you take a system from a sketch to production and keep improving it.
- Translator: you can sit with a business user, understand how they make a decision, and turn it into logic a system can run and test.
- Pragmatic: you ship a simple first version quickly, reuse what already works, and add complexity only when the data and feedback justify it.
- Honest about accuracy: you test your own work before anyone else does and say so when something does not work.
- People-first view of AI: you believe AI should make people stronger, and you measure success by adoption and outcomes, not features shipped.
Solen-isms (How We Operate)
- Ownership Mindset: We own the mission. We don't wait to be told. We lead, fix, and build as if it were ours. When problems arise, we run toward them, not away.
- Benchmark to Best: We seek out excellence. We measure ourselves against the best and adapt quickly, always looking for the edge that moves us forward.
- Openly Self-Critical: Leaders go first: they surface their misses early, ask for direct feedback, and turn “game film” into better decisions and better systems.
- Disagree & Commit: Leaders disagree openly and constructively, regardless of title or tenure, because truth is how we avoid bureaucracy and benchmark to best.
- Figureitoutiveness: Embracing challenges, fostering growth, and taking initiative, leaders thrive on hard work and high failure tolerance.
- Mutual Elevation: Committed to growth, elevating teams, developing future leaders, and removing barriers to mastery and joy.
- Learn-it-alls: Learn-it-alls, not know-it-alls, embracing failure as part of growth and progress.
- Pioneering: Pioneering driven by opportunity, using first principles thinking and unbound by tradition or rules.
- Will to Win: Leaders relentlessly pursue potential, reject mediocrity, and get what they tolerate.
- Dialectical: We learn through conversation. We ask, listen, question, and evolve - valuing disagreement as a tool for alignment and deeper understanding.
- Frugality: Achieving more with less, focusing on quality, long-term value, and company-wide responsibility.
- Candor: We communicate with integrity, transparency, and respect in all conversations.
- Kaizen: Kaizen is continuous improvement as a way of life. Leaders make hundreds of small upgrades that compound, turning lessons into standards and standards into durable systems.
Recruitment Disclaimer
Solen Software Group will never request payment, banking information, or other sensitive personal or financial details during any stage of the recruitment process. All official communication will come from a verified @solensoftwaregroup.com email address. Any outreach outside these standards should be disregarded and reported through our official channels.
We thank all applicants for their interest in joining Solen Software Group. Only candidates selected for an interview will be contacted.

